<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[LoopGIF]]></title><description><![CDATA[LoopGIF]]></description><link>https://loopgif.hashnode.dev</link><generator>RSS for Node</generator><lastBuildDate>Wed, 18 Nov 2026 15:05:49 GMT</lastBuildDate><atom:link href="https://loopgif.hashnode.dev/rss.xml" rel="self" type="application/rss+xml"/><language><![CDATA[en]]></language><ttl>60</ttl><item><title><![CDATA[A Twitch Emote Needs One Square GIF or Three Exact Sizes, Not the Raw Clip You Recorded]]></title><description><![CDATA[You already have the moment: a few seconds of a stream, a face you want as a reaction, a sticker, or a short word for chat. You export a clip or save a GIF…]]></description><link>https://loopgif.hashnode.dev/a-twitch-emote-needs-one-square-gif-or-three-exact-sizes-not-the-raw-clip-you-recorded</link><guid isPermaLink="true">https://loopgif.hashnode.dev/a-twitch-emote-needs-one-square-gif-or-three-exact-sizes-not-the-raw-clip-you-recorded</guid><category><![CDATA[gif]]></category><category><![CDATA[animation]]></category><category><![CDATA[Design]]></category><category><![CDATA[Tutorial]]></category><category><![CDATA[webdev]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 14:19:54 GMT</pubDate><content:encoded><![CDATA[<p>You already have the moment: a few seconds of a stream, a face you want as a reaction, a sticker, or a short word for chat. You export a clip or save a GIF and send it to Twitch as an emote. The upload comes back. The file is not square, it has too many frames, it is over the size cap, or it flashes when it loops. The joke was fine. The file was not an emote yet.</p>
<p><em>I build LoopGIF.</em></p>
<p>An animated Twitch emote, on the limits the tool’s page checked on 7 November 2026 against Twitch’s Animated Emotes and emote formatting guidelines, is a square GIF. Twitch describes two ways to upload one. In auto-resize mode you upload a single square GIF between 112 × 112 and 4096 × 4096 pixels, up to 1 MB, and Twitch makes the smaller sizes. In manual mode you upload three GIFs yourself, at 28 × 28, 56 × 56 and 112 × 112 pixels, each up to 512 KB. The work is to crop a square, keep the motion under those limits, and look at the file before you submit it. Those numbers can change, so read Twitch’s current guidelines before you upload. Animated-emote eligibility depends on the channel’s status; a file tool only prepares the file. LoopGIF is not affiliated with, endorsed by or sponsored by Twitch, Discord or Slack. Their names are used only to say which service a file is made for.</p>
<h2>Auto-resize wants one square GIF; manual upload wants three exact sizes</h2>
<p>Auto-resize is one square GIF in that 112-to-4096 range, no larger than 1 MB; Twitch builds the versions that appear at the smaller chat sizes. Manual upload is stricter about pixels: you provide all three sizes yourself, and each file has to be exactly 28, 56 or 112 pixels on a side and no larger than 512 KB.</p>
<p>Pick one GIF when you are happy to let Twitch make the small versions. Pick the three-file ZIP when those small versions matter. A 28 pixel emote needs bolder shapes than a 112 pixel one.</p>
<p>Twitch writes 1 MB and 512 KB without saying whether a KB is 1,000 or 1,024 bytes. The checks on this page use 1,000, which is the stricter reading, and the checklist shows sizes the same way.</p>
<p>A 375 KB GIF at 320 × 320 pixels with 16 frames, cropped as a full square and resized to the default 112 × 112 with 255 colours, becomes a 71.2 KB, 16-frame GIF — a small part of the 1,000 KB auto-resize allowance. A 503 KB source at the same size and frame count, resized to 224 × 224 with its transparency kept, comes out at 276 KB and 16 frames, still under the single-file cap. The tool offers 112 × 112 by default, and 224 and 448 when your cropped square is big enough. Manual mode packs three files in a ZIP. In one example, a 510 KB, 320 × 320, 16-frame GIF becomes a 4.7 KB file at 28 pixels, a 16.6 KB file at 56 pixels, and a 57.5 KB file at 112 pixels; the ZIP is 79.1 KB, and each GIF sits well under its 512 KB limit.</p>
<h2>The square, the background, and the first frame</h2>
<p>A widescreen clip of a stream is not square. On this tool you drag a square box over the part you want to keep, or type its X, Y and size.</p>
<p>If the clip sits on a plain colour, you can tick Remove plain background and adjust the tolerance. That control clears the background colour that touches the edge of the square, and it works on plain or near-plain backgrounds only. GIF transparency is on or off, with no soft edges. Soft edges blend toward the edge colour you pick, so a fuzzy halo is not a see-through edge.</p>
<p>Then look at the first frame. Twitch uses it as the still thumbnail unless you set a custom one. A frame in the middle of a blink or a wave makes a poor thumbnail. If the clip starts on an awkward pose, start it on a good frame before you make the emote.</p>
<h2>Sixty frames, and the cap on each file</h2>
<p>Twitch allows at most 60 frames in an animated emote. When a clip has more, this tool keeps every second, third or fourth frame — whatever it takes to reach 60 or fewer — and adds the delays of the dropped frames to the frame before. The emote still lasts as long as the original. It just plays with fewer, longer frames.</p>
<p>One example is a sticker repeated four times: a 299 KB GIF at 160 × 160 pixels with 64 frames, already over the limit. After a 160 pixel square crop, a resize to 112 × 112 with 255 colours, and a pass that keeps one frame in every two and merges the delays, the result is a 131 KB, 112 × 112 GIF with 32 frames, inside the 1,000 KB cap, with the same duration as the long source.</p>
<p>File size is a separate ceiling: 1 MB for the one auto-resize GIF, and 512 KB for each of the three manual files. The tool tries stronger settings until each file fits, and it says so if one cannot.</p>
<h2>The flashing estimate is a hint, not a review</h2>
<p>Twitch says an animated emote must not flash or flicker more than three times in one second. A black-and-white strobe will trip that rule. A calm bounce or a slow glow will not. Everything in between is a judgement call, and a tool cannot know how Twitch will judge your emote.</p>
<p>What this page can do is warn you. It measures the average brightness of every frame, counts each jump of more than a fifth of the full brightness range, up or down, including the jump from the last frame back to the first, and reports the most jumps it finds in any one second of the looping GIF. More than three in a second triggers the warning. This is the tool’s own estimate, not Twitch’s review, and counting every jump rather than every full flash is deliberately strict. Treat the result as a hint to soften a flicker.</p>
<p>The Text emoji example types the word “LOL,” puts a shake on a transparent square, and resizes it to 112 × 112 with 255 colours. The result is a 10.0 KB GIF with 12 frames. The shake moves the letters but hardly changes the brightness, so the flashing estimate stays quiet.</p>
<h2>Making the file, then reading the checklist</h2>
<p>I use <a href="https://loopgif.net/twitch-emote-maker">LoopGIF’s Twitch emote maker</a> for this path. You can drop a GIF, a short video or a picture on the Upload tab, or switch to Text emoji and type a short word. Accepted file types are GIF, PNG, JPG, WebP, MP4, MOV or WebM. GIFs and pictures can be up to 25.0 MB. The file is opened on your device and never uploaded. The tool is free, needs no sign-up, runs in your browser, adds no watermark, and needs no account. A separate Generate with AI tab costs 1 credit per sticker and needs a signed-in account; that is not this path.</p>
<p>Type an emote name for the file names — lower case, with letters, numbers and underscores only. Frame the square. Then choose One GIF for auto-resize and a size, Three GIFs in a ZIP for manual upload, or Still PNG. Press Make emote. The checklist shows size, dimensions, frames and the flashing estimate, and the previews show the emote at 28, 56 and 112 pixels. Then you download the GIF or the ZIP.</p>
<p>Uploads usually fail when the file is not square, is over the size limit, has more than 60 frames, flashes, or, in manual mode, is not exactly 28, 56 and 112 pixels. The checklist covers the first four of those, and the ZIP is built with exactly those three sizes. Eligibility and Twitch’s own review can still turn an emote down, and a file tool cannot see either. Treat the checklist as a pre-flight check, not as a promise of approval.</p>
<p>This tool can save a still PNG at 28, 56 or 112 pixels, but it does not check a size limit, because the page did not verify one for static emotes. It was built and tested in current desktop Chrome, relying on GIF files, background encoding with Web Workers, and video decoding with WebCodecs, which the page checks live. It has not been tested in every browser.</p>
<h2>When a Twitch-ready file is the thing you actually need</h2>
<p>A channel emote from a clip is a short moment from a stream recording, turned into a looping square for auto-resize or a three-file ZIP. A reaction for chat is a tighter crop: a character or a face, with the edge left transparent when the background is plain enough to clear. A word as an emote does not need a source file; you type a short word such as LOL and let it bounce, wave or shake on a transparent square. Three sizes for manual upload give you the 28, 56 and 112 pixel versions, each checked against its own 512 KB cap, in one ZIP. A calmer version of a flashy GIF lets you see how much the clip flickers in its busiest second and decide whether to soften it before you submit. In all of these, the destination is Twitch’s upload screen, not a generic resize for some other service.</p>
<h2>What the tool does, and where it stops</h2>
<p>The source can be a GIF, a short video, a picture, or a typed word, opened in the browser: GIFs and pictures up to 25.0 MB, with no upload, no account, and no watermark. The output is one square GIF for auto-resize, three GIFs at 28, 56 and 112 pixels in a ZIP, or a still PNG at one of those sizes. Animated files are kept to 60 frames or fewer by dropping frames evenly and merging delays, and the tool tries to fit each GIF under its size limit. The checklist reports size, dimensions, frames and a flashing estimate, which is the tool’s own, not Twitch’s review.</p>
<p>The platform limits were checked on 7 November 2026 and can change. Eligibility depends on the channel’s status. Background removal works only on plain or near-plain backgrounds, and GIF transparency has no soft edges. You go in knowing that a raw clip is not yet an emote, and you come out with a square GIF, or a ZIP of three exact sizes, that you can hold up against Twitch’s current guidelines.</p>
<p><em>I build LoopGIF.</em></p>
]]></content:encoded></item><item><title><![CDATA[A Transparent GIF Saved as JPG Needs a Fill Colour, or Many Programs Show Black]]></title><description><![CDATA[You already have a GIF that plays: a sticker with a cut-out edge, a meme, a short product loop. Then you need one still. A forum or a form will take a JPG…]]></description><link>https://loopgif.hashnode.dev/a-transparent-gif-saved-as-jpg-needs-a-fill-colour-or-many-programs-show-black</link><guid isPermaLink="true">https://loopgif.hashnode.dev/a-transparent-gif-saved-as-jpg-needs-a-fill-colour-or-many-programs-show-black</guid><category><![CDATA[gif]]></category><category><![CDATA[animation]]></category><category><![CDATA[Design]]></category><category><![CDATA[Tutorial]]></category><category><![CDATA[webdev]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 14:11:46 GMT</pubDate><content:encoded><![CDATA[<p>You already have a GIF that plays: a sticker with a cut-out edge, a meme, a short product loop. Then you need one still. A forum or a form will take a JPG and not a GIF. An email or a document will not play the animation. A printer or a photo site expects a normal picture with a solid background. The tempting step is to convert the file and hope the frame survives. When the GIF was transparent, the result often looks wrong. The subject sits in a black box, or the cut-out area fills with a colour you never chose.</p>
<p><em>I build LoopGIF.</em></p>
<p>That black box is not a stain in the original file. JPG cannot hold transparency. Every pixel has to have a colour. A transparent GIF saved with no background would show black in many programs, which is almost never what you want. The work, then, is not turning a motion file into a photograph by magic. It is choosing one frame, painting that frame onto a colour, and saving the canvas as a JPEG. The file that comes out is a normal .jpg — JPG and JPEG are the same format, and .jpg is the shorter, more common spelling — and it opens anywhere. It is a still. It does not animate. If you still need the motion, keep the GIF.</p>
<h2>Every see-through pixel has to become a colour</h2>
<p>A GIF can leave pixels empty so the page, the slide, or the chat background shows through. A JPG cannot do that. Once you ask for a JPEG, the hole in the picture has to be filled, and the fill becomes part of the image. There is no later switch that makes those pixels empty again, short of editing the file by hand and hoping the subject still separates cleanly from the painted ground.</p>
<p>If the GIF has no transparency, the fill does not matter. Every pixel is already opaque, and the still is simply that frame, encoded as a JPEG. The converter’s own examples include a meme-style frame saved this way: frame 12, background #ffffff, quality 90, from a 719 KB GIF at 320 × 320 pixels to a 24 KB JPG of the same size. The white was there in the settings, but it had nothing to paint over.</p>
<p>If the GIF is a sticker, the fill is the whole point. White is the default, and it suits most pages and documents. On a white field, a transparent sticker starts to look like a normal photo-style picture. One published example takes frame 7 of a 276 KB, 320 × 320 sticker, paints it on #ffffff at quality 85, and gets an 18 KB JPG. The same kind of sticker on a dark background tells a different story. Frame 5 of a 349 KB file, painted on #17141f at quality 90, becomes a 24 KB JPG that would sit more naturally on a dark page or a dark profile. The useful habit is to choose a background that matches where the picture will be used.</p>
<p>You can pick that colour with a colour field, or with swatches for white, black, or a warm paper tone. A preview that already shows the background is worth more than a guess, because the colour you see is the colour that is saved. Try a few, then commit. After the save, the background is solid. Re-opening the JPG, editing it, and saving it again is a lossy trip: a little more detail disappears each time.</p>
<p>If what you actually need is the hole in the background, a JPG is the wrong output. That is a different job, for a format that can stay see-through. This article is about the case where a solid picture is what the destination wants.</p>
<h2>Quality is how you trade sharpness for a smaller file</h2>
<p>JPG is lossy. It saves space by discarding detail the eye is less likely to miss. The quality control on this kind of converter runs from 50 to 100. Around 85 to 90 is hard to tell from the original on most pictures, and that is the range to use when you care about the still. Lower numbers give smaller files and softer detail. A setting of 60 is visibly softer but much smaller, and it is fine for a preview or a thumbnail if you can live with the compression when you zoom in on the edges. One example does exactly that: frame 9, a white background, quality 60, from a 768 KB, 320 × 320 GIF to an 18 KB JPG.</p>
<p>The same guidance also suggests quality 70 to 80 when the still is only a thumbnail, and 60 to 70 for previews and thumbnails more generally. Those ranges overlap on purpose. A preview can be a little softer. A picture you will print, or place as a product or profile image, should stay higher.</p>
<p>The kind of picture matters as much as the number. Flat artwork with sharp edges, such as cartoon stickers, shows JPEG blur more than photographs do, so keep the quality higher for it. A lower setting gives a smaller file with some blur around edges, which is exactly what a hard outline will show.</p>
<p>After you save, the size shown is the real file size. If it is bigger than you want, lower the quality a notch or choose a smaller size from the Size menu and save again. Both are quick. There is also a hard ceiling: the JPG is limited to 4096 pixels on the longer side.</p>
<h2>Choose the frame, then save one still</h2>
<p>“Convert a GIF to a JPG” is easy to over-read. The file has many moments. The JPG can hold only one. You scrub to the frame you want, check it in a preview that already includes the background colour, and then make the file.</p>
<p>I use <a href="https://loopgif.net/gif-to-jpg">LoopGIF’s GIF to JPG tool</a> for that path. You drop a GIF onto the box or click to choose one. It has to be a GIF, up to 25 MB and 300 frames, and it is opened in your browser. The tool is free, needs no sign-up, and uploads nothing. Your file never leaves your device. The frame is drawn onto a canvas filled with your background colour and encoded there, so the GIF stays on your computer or phone. You do not need an account. The page’s own description of the app also states that there is no watermark.</p>
<p>Pick the frame with the slider or the arrow buttons. Set the colour and the quality. Press Save this frame as JPG. The result appears with its size, and Download saves a file named like your GIF with the frame number. That is one frame per press. If you need a still of a later moment, you choose that frame and save again. Exporting a whole sequence as a pack of files is a different job. This page does not do it.</p>
<p>The tool was built and tested in current Chrome. I am not going to guess at other browsers from their names. The useful guarantee is local: the colour in the preview is the colour in the file.</p>
<h2>When a solid still is the thing you actually need</h2>
<p>A forum or a form that rejects GIF is the obvious case. Some upload fields take JPG and not GIF. Save a frame and upload that.</p>
<p>A small preview image is another. JPG files are small, which is the point of the format, and a frame at a mid-range quality is enough for a thumbnail.</p>
<p>An email or a document is a third. Drop a still of the moment into a message where an animation would not play, and the reader sees the pose you chose instead of a first frame or a broken attachment.</p>
<p>A print-friendly picture is a fourth. JPG is the format printers and photo sites expect, with a solid background, so you are not asking a press or a lab to invent a colour for empty pixels.</p>
<p>A product or profile picture is a fifth. Take one good frame of an animated sticker and place it on a colour of your choice, then use that still where a loop will not run.</p>
<p>In all of these, the GIF remains the moving original. The JPG is a still, and it should be treated as one. If the destination needed a see-through edge, you would not be here. If it needed every frame, you would be packing stills, not saving one. The job is narrower, and that is what makes it reliable: one frame, a fill you can see before you commit, a quality setting that matches how closely the still will be looked at, and a file that ordinary software already knows how to open.</p>
<h2>What the tool does, and where it stops</h2>
<p>This is the shape of the work, and the line around it.</p>
<p>The file is accepted only as a GIF, up to 25 MB and 300 frames, and is opened in the browser. Nothing is uploaded, and no account is required. The JPG is limited to 4096 pixels on the longer side.</p>
<p>JPG has no transparency. Every transparent pixel becomes the background colour you pick, so the picture will have a solid background. White is the default. The preview updates as you change the colour.</p>
<p>Quality runs from 50 to 100. The JPG is lossy, so saving a frame as JPG and then editing and re-saving it loses a little more each time.</p>
<p>One frame is written per press. The file name follows the GIF and includes the frame number. The result does not animate.</p>
<p>Those limits are the point, not a footnote. You go in knowing that the still will be opaque, that the fill is a choice you have to make, and that sharpness and file size pull against each other. You come out with a .jpg you can actually hand to a form, a message, a printer, or a profile, instead of a transparent GIF that those places cannot use, or a conversion that quietly filled the holes with black.</p>
<p><em>I build LoopGIF.</em></p>
]]></content:encoded></item><item><title><![CDATA[A Screenshot of a GIF Keeps the Checkerboard; a PNG of the Frame Does Not]]></title><description><![CDATA[A looping sticker, a bouncing logo, or a short interface clip arrives as a GIF, and then you need one still: a thumbnail, an icon, a slide image, or a…]]></description><link>https://loopgif.hashnode.dev/a-screenshot-of-a-gif-keeps-the-checkerboard-a-png-of-the-frame-does-not</link><guid isPermaLink="true">https://loopgif.hashnode.dev/a-screenshot-of-a-gif-keeps-the-checkerboard-a-png-of-the-frame-does-not</guid><category><![CDATA[gif]]></category><category><![CDATA[animation]]></category><category><![CDATA[Design]]></category><category><![CDATA[Tutorial]]></category><category><![CDATA[webdev]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 14:08:26 GMT</pubDate><content:encoded><![CDATA[<p>A looping sticker, a bouncing logo, or a short interface clip arrives as a GIF, and then you need one still: a thumbnail, an icon, a slide image, or a frame for an editor. The usual shortcut is a screenshot, which often keeps the checkerboard that was only there to show transparency, or a scrap of the player around the picture. A format without transparency, such as JPG, would fill the background with a colour and put a box around the subject. What you actually want is the frame itself, as a PNG, with the see-through pixels still see-through.</p>
<p><em>I build LoopGIF.</em></p>
<h2>The checkerboard is a preview, not part of the picture</h2>
<p>When a GIF has transparent pixels, a decent preview draws them on a checkerboard so you can tell the background is empty. That board is a viewing aid. It is not in the file. A screenshot captures what is on the screen: the board, and sometimes the chrome of the tab or the player. Placed on a slide or a website, that grab leaves the subject on a chequered ground.</p>
<p>The useful job is to capture the exact frame without a screenshot tool and without the checkerboard. A tool built for this shows the frame at full detail on a checkerboard so the transparent areas are easy to see, then writes a PNG that keeps those areas empty.</p>
<p>A PNG shows the frame, not the animation, and it cannot play back. If you still need the motion, the GIF stays a GIF, and the PNG sits beside it as a poster, a thumbnail, or a reference.</p>
<h2>The PNG is the picture a browser would draw, including earlier frames</h2>
<p>A GIF does not always store each frame as a complete picture. Many files only write the region that changed, and rely on earlier frames to fill in the rest. If you pulled the raw block for frame ten, you might get a patch, not the whole sticker. The still you want is the picture as a browser would draw it at that moment, with earlier frames composited in.</p>
<p>That is why a single saved frame can be a complete bear holding a heart, or a logo mid-bounce, rather than a fragment of ink. The compositing is the same drawing a browser already does when it plays the loop. The PNG just keeps that drawing still.</p>
<p>If the file is already a still image, or it has only one frame, there is nothing to scrub, and the PNG is that frame. For a loop, you choose the moment: the first frame, where the slider starts, a later one that is clearer for a thumbnail, or a pose that works as an icon.</p>
<p>“Convert a GIF to a PNG” is easy to over-read. This page is not writing a numbered set of files into a ZIP, and it is not packing every frame into one grid. It saves one frame at a time. A ZIP of every frame, or a selection, and a sprite sheet that holds every frame in one image, are different jobs on the same site. The work here is one moment, one file.</p>
<h2>Transparency is kept, with hard edges</h2>
<p>A GIF can have see-through pixels, and PNG is the common format that can keep them. Saved that way, the background stays see-through, so you can place the picture on a slide, a website, or an image editor without a box around it.</p>
<p>GIF transparency is on or off for each pixel, so edges are hard. The PNG keeps exactly what the frame shows, hard edges included. That is faithful, and it can surprise you if you hoped a conversion would smooth a jagged outline. The conversion does not invent partial transparency the GIF never had.</p>
<p>Sometimes you would rather have a coloured ground, because the still is going onto a surface that cannot show transparency, or because a solid colour is part of the design. Tick the fill option and choose a colour. The option is off by default.</p>
<h2>Scrub to the frame, then choose how the PNG is saved</h2>
<p>The practical path is short, and I use <a href="https://loopgif.net/gif-to-png">LoopGIF’s GIF to PNG tool</a> for it. You drop a GIF onto the box or click to choose one. The file can be up to 25 MB and 300 frames, and it is opened on your device. Nothing is uploaded, and you need no account.</p>
<p>After the GIF is open, scrub to the frame you want. Drag the frame slider, or use the arrow buttons, which step one frame at a time and are the quickest way to land on an exact frame. The slider also responds to the arrow keys. The preview shows the frame at full detail on a checkerboard, and the line below gives the frame number and the time. For the first frame, load the GIF and leave the slider at frame 1, which is where it starts.</p>
<p>Then choose the PNG options. Transparency is kept by default. You can fill transparent areas with a colour instead, or pick a smaller size such as 480 or 320 px wide. If the GIF is heavy and you only want a small picture of it, shrink the still with the Size menu before you download. A 480 px wide PNG is much smaller than a full-size one and is enough for most previews. The PNG can be as large as the GIF itself, up to 4096 px on the longer side.</p>
<p>Press “Save this frame as PNG.” The picture appears beside the frame with its size, and Download saves a file named like your GIF, with the frame number. Saving several frames of the same GIF, one press at a time, gives files that sort in order. Each press still makes one PNG. That is a convenient way to take two or three poses, not a batch export of the whole loop.</p>
<h2>Why a PNG can be larger than the GIF, and when it is smaller</h2>
<p>A still is often lighter than the animation it came from, because you are keeping one frame rather than the whole loop. It can also be larger. A PNG of a full frame stores every pixel with its own colour and optional transparency. A GIF stores indexed colours across all frames, with a lot of sharing. One complete, lossless still can weigh more than a compact, shared palette that had to describe many frames at once.</p>
<p>The way to cut a large PNG is to choose a smaller size. The page’s own examples, made from its own GIFs and drawn the way the tool draws a frame, show how size and subject change the result.</p>
<p>One example saves frame 10 of a 16-frame animated sticker — a cartoon bear holding a heart — at the GIF’s full 320 × 320 px, with transparency kept. The GIF is 356 KB; the PNG is 54 KB. The checkerboard behind the animation is the transparency, and the PNG keeps it, so the bear can sit on any background. A second example takes the same idea to 160 px wide: frame 6 of a 16-frame smiling sun, from a 320 × 320 px GIF of 491 KB, becomes a 160 × 160 px PNG of 34 KB, which suits an icon. A third is a text GIF with a clear background: frame 8 of 12, a glitter frame at 400 × 200 px, goes from a 62 KB GIF to a 12 KB PNG with the sparkles and the letters and nothing behind them. A fourth is a bakery logo mid-bounce: frame 12 of 23, saved at 240 × 240 px from a 320 × 320 px, 168 KB GIF, becomes a 19 KB PNG. A single frame of a loop is a quick way to get a still version of a logo mid-motion.</p>
<p>Those figures are the page’s examples, not a promise that every GIF will shrink by the same ratio. If your PNG is larger than the GIF, a smaller size in the Size menu is the lever the page gives you.</p>
<h2>When one still is the right output</h2>
<p>The cases are ordinary, and easy to blur with jobs this page does not do. A thumbnail uses the clearest frame as the preview image on a page or in a document, as a real picture rather than a screenshot of a player, often with the background left empty so it does not arrive with a box. A transparent sticker or icon keeps one pose of an animated sticker as a still with no background: you are choosing a pose, not flattening the loop. A frame for a slide or a doc puts the moment in a deck, where animation would be distracting, and in documents that will accept a PNG but will not play a loop well. An image editor can open the PNG to retouch a frame, or use it as a reference, which a screenshot that baked in a checkerboard will not support cleanly. And a screenshot of an animation, in the useful sense, is just the exact frame, without a screenshot tool and without the checkerboard.</p>
<p>All of those are one frame. If what you needed was every frame as a numbered set, a flat image with a chosen background colour and a quality setting, or a resize of the animation itself, those are other tools. This page stops at the still.</p>
<h2>What the tool does, and where it stops</h2>
<p>The work runs in your browser. Your GIF never leaves your device. It is opened, the frame is drawn and saved, and nothing is sent to a server. You need no account and no sign-up. The tool is free, and it does not add a watermark.</p>
<p>The limits are on the page. GIFs up to 25 MB and 300 frames are accepted. The PNG can be as large as the GIF itself, up to 4096 px on the longer side. One frame is written per press. The PNG is the frame as a browser draws it, with earlier frames composited in, and with GIF’s hard, on-or-off transparency. The PNG cannot play the animation back.</p>
<p>The tool was built and tested in current Chrome. GIFs open everywhere, using your browser’s decoder or the tool’s own. I am not going to list other browsers from a live support checklist that, on a plain fetch of the page, only shows a “GIF files” row marked “checking.”</p>
<p>This page does not write a ZIP of every frame, turn the loop into a different moving format, or edit or convert video. It takes the GIF you have, lets you stop on a frame, and gives you a PNG of that moment.</p>
<p><em>I build LoopGIF.</em></p>
]]></content:encoded></item><item><title><![CDATA[A Photo-Like GIF Shrinks Most as WebP; a Flat, Busy Banner Often Barely Shrinks]]></title><description><![CDATA[You already have a GIF that plays. It might be a loop on a web page, a sticker whose edges should look cut out, or a photo-like animation that has grown…]]></description><link>https://loopgif.hashnode.dev/a-photo-like-gif-shrinks-most-as-webp-a-flat-busy-banner-often-barely-shrinks</link><guid isPermaLink="true">https://loopgif.hashnode.dev/a-photo-like-gif-shrinks-most-as-webp-a-flat-busy-banner-often-barely-shrinks</guid><category><![CDATA[gif]]></category><category><![CDATA[Design]]></category><category><![CDATA[webdev]]></category><category><![CDATA[animation]]></category><category><![CDATA[Tutorial]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 14:04:15 GMT</pubDate><content:encoded><![CDATA[<p>You already have a GIF that plays. It might be a loop on a web page, a sticker whose edges should look cut out, or a photo-like animation that has grown heavier than you expected. GIF can use at most 256 colours in each frame, and a pixel is either fully see-through or fully opaque. Current browsers can show an animated WebP in an ordinary image tag. The work is not inventing a new loop. It is asking whether the same frames can travel in a file that is often lighter, and whether the edges can stay soft.</p>
<p><em>I build LoopGIF.</em></p>
<p>That change of format is easy to oversell. WebP is often smaller than GIF, because it compresses photographic detail and gradients far better than a paletted frame can. It also has a real alpha channel, so an edge can be partly see-through instead of on or off. None of that is a guarantee. A flat, simple GIF can already be tiny, and a high quality setting can make the WebP larger than the GIF it came from. The only honest check is to look at both sizes for the file you actually have, then lower the quality or the width, or keep the GIF.</p>
<h2>Some pictures shrink a lot; some barely move</h2>
<p>The difference is in the picture, not in the file extension. Glow and soft gradients are what GIF handles worst and lossy WebP handles best, so that is where a file shrinks the most. Flat colours with hard edges are already friendly to GIF, so the saving there is often modest.</p>
<p>The converter’s own page shows this with four real conversions, made from GIFs already used on the site, by the same writer the tool uses. The frame images come from the Skia library’s WebP encoder in the site’s script; your browser uses its own encoder. The numbers on the cards are the real ones for those files, not a prediction of the exact size your browser will write.</p>
<p>A 12-frame neon sign, converted at quality 75 and 480 pixels wide, with no transparency, goes from a 130 KB GIF at 480 × 270 to a 33 KB WebP of the same dimensions. A sticker GIF with a transparent background, converted at quality 80 and 320 pixels wide, with transparency kept, goes from 356 KB at 320 × 320 to 161 KB; the transparency is stored in the WebP’s own alpha data, so the edges are soft instead of on or off. A third example trades more detail for size: quality 50 and a width of 240 pixels instead of 320 takes a 320 KB GIF at 320 × 320 down to a 91 KB WebP at 240 × 240, the lightest of the four, with a reminder to look at the caption text and judge whether it is still readable enough. A scrolling banner at quality 70 and 480 pixels wide only goes from 215 KB at 480 × 120 to 183 KB, because the motion is almost a full-frame change every time and the colours are already the kind GIF encodes well. If your file looks like the neon sign or the sticker, the format change is likely worth doing. If it looks like the banner, you may still want WebP for the edges, but you should not expect a dramatic saving.</p>
<h2>Only the changed part of each frame is written</h2>
<p>An animated WebP is a container with a header that says the canvas size and whether there is transparency, a block that sets the loop, and one block for each frame. Each frame block holds the frame’s position, its size, how long it shows, and the image data. The page writes that container itself, in the browser, and asks your browser’s encoder only for the image data of each frame.</p>
<p>The frame image can come back in three shapes: a lossy picture, a lossless one, or a lossy picture with a separate transparency layer. The writer reads all three and puts the pieces into the frame block. Positions in a WebP are stored in steps of two pixels, so every frame starts on an even pixel, and each duration is stored in milliseconds. Frame timing is kept in milliseconds as the GIF stored it, and the loop is set to repeat forever. You get an animated WebP with the same frames and timing.</p>
<p>After the first frame, the tool compares each frame with the one before it and finds the smallest rectangle that contains every changed pixel. Only that rectangle is encoded, and the player puts it on top of what is already on screen. A bouncing object on a still background means small rectangles and a small file. A frame that changes nothing is not written at all; its time is added to the frame before it, so the animation lasts exactly as long as the GIF did, and the result tells you when this happened. A scrolling banner or a video-like GIF, which changes the whole picture every frame, has little to skip. This is still a conversion of a GIF you already have, not a way to assemble still photographs or to turn a video into an animated image.</p>
<h2>Soft edges can stay soft, because WebP has a real alpha channel</h2>
<p>Transparency is the other reason to leave GIF, even when the size saving is modest. In a GIF, a pixel is either fully transparent or fully opaque. WebP keeps partial transparency, so edges look smoother than that on-or-off mask. If the GIF has see-through pixels, they are kept by default. A WebP has a real alpha channel, so transparent pixels stay transparent. You can switch it off and fill the background with a colour of your choice instead.</p>
<p>That matters for a sticker or any cut-out that will sit on more than one background. The page’s sticker example keeps transparency on the same 320 × 320 animation, with edges stored as alpha rather than as a hard cut. If you would rather have a solid background, fill it in at convert time. If the place you are uploading to still asks for GIF, or an older app will only show the first frame of an animated WebP, keep the GIF.</p>
<h2>Quality and width are the levers, and the sizes have to be compared</h2>
<p>The quality slider runs from 1 to 100, and the default on the page is 75. The value is passed to your browser’s WebP encoder. Lower numbers make a smaller file with softer detail, and lower values can also produce more blocky areas, so check the result by eye before you use it.</p>
<p>The top of the slider is not a single behaviour in every browser. In Chrome, quality 100 writes lossless frames, and for a GIF with few colours that can come out about as large as the GIF. Other browsers may treat the top of the slider differently. That is one way a WebP can fail to shrink, and a reason to read both file sizes instead of assuming the conversion worked because it finished.</p>
<p>Width is the other lever. You can keep the original width or pick a smaller one. The line under the width shows the size of the result. The third example on the page uses both levers at once, quality 50 and 240 pixels instead of 320, and it is the lightest of the four for that reason. If the WebP is not smaller, lower the quality or the width, then compare the sizes again, or keep the GIF. The page always shows the real numbers for your file, before and after, with the frame count on the result. Move the quality slider, convert again, and compare the real sizes of each attempt. A setting that looks fine on a neon glow may be too soft on caption text.</p>
<h2>How the conversion runs, and when to keep the GIF</h2>
<p>The practical path is short, and I use <a href="https://loopgif.net/gif-to-webp">LoopGIF’s GIF to WebP tool</a> for it. You drop a .gif file onto the box or click to choose one. Files up to 25 MB, 300 frames, and 30 seconds of playing time are accepted. After you set quality, width, and transparency, press Convert to WebP. The WebP appears beside the GIF with both file sizes and the frame count, and Download saves it.</p>
<p>All of that happens in the browser. Your file never leaves your device. The GIF is decoded and the WebP is written in your browser, with no upload and no account. The converter is free, does not require a sign-up, and does not add a watermark.</p>
<p>Each frame is encoded by your browser’s own WebP encoder. If it hands back another format, the page says your browser cannot encode WebP and stops, rather than saving a wrong file. Some browsers return a PNG when asked for WebP. The tool was built and tested in current Chrome. The page also checks the browser you are using right now, by asking it directly rather than guessing from the name, for WebP encoding and for GIF files. If the check fails, try Chrome, or another browser that can save canvas images as WebP.</p>
<p>Animated WebP is shown by current browsers, including in an image tag on a web page, but some older apps and tools only show its first frame. If the place you are uploading to still asks for GIF, keep the GIF. This converter does not turn a WebP back into a GIF, join still photographs, or start from a video. It takes an existing GIF and writes an animated WebP.</p>
<h2>When the format change is the right job</h2>
<p>The useful cases are ordinary. On a web page with animated images, current browsers show animated WebP in an image tag, and the file is often lighter than the GIF. For a sticker with soft edges, WebP keeps partial transparency, so the cut-out looks smoother than a GIF’s on-or-off mask. For a photo-like GIF that is too heavy, gradients and photographic frames are where the format change pays off most.</p>
<p>Go in knowing the limits. The frames and their timing come across, with durations stored in milliseconds, and identical frames merged into the frame before them. The loop will repeat forever. Soft edges can stay soft if you keep transparency. The file will often be smaller, especially when the GIF was spending its budget on glow, gradients, or photographic detail. It will not always be smaller, especially when the GIF was already flat or the quality setting is high, and the only way to know is to look at both sizes. If an older app will not play the animation, the lighter file is not the useful one.</p>
<p><em>I build LoopGIF.</em></p>
]]></content:encoded></item><item><title><![CDATA[An Animated WebP Keeps Its Timing in a GIF, but Size, Colour, and Soft Edges Change]]></title><description><![CDATA[You export a sticker, a flickering sign, or a short loop as WebP because the file is small and the edges look clean. Then you try to attach it to an email…]]></description><link>https://loopgif.hashnode.dev/an-animated-webp-keeps-its-timing-in-a-gif-but-size-colour-and-soft-edges-change</link><guid isPermaLink="true">https://loopgif.hashnode.dev/an-animated-webp-keeps-its-timing-in-a-gif-but-size-colour-and-soft-edges-change</guid><category><![CDATA[gif]]></category><category><![CDATA[animation]]></category><category><![CDATA[Design]]></category><category><![CDATA[Tutorial]]></category><category><![CDATA[webdev]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 13:58:51 GMT</pubDate><content:encoded><![CDATA[<p>You export a sticker, a flickering sign, or a short loop as WebP because the file is small and the edges look clean. Then you try to attach it to an email, drop it into a forum that only lists GIF, or hand it to a chat app or an older editor. The motion is already in the file. The file just will not play there, or the form will not take the type. The work is not inventing an animation. It is translating one you already have into a format more places still accept.</p>
<p><em>I build LoopGIF.</em></p>
<p>That translation has a cost, and the cost is easy to misread as a failed encode. WebP can be lossy or lossless, it can hold a full alpha channel, and it usually makes a much smaller file than a GIF of the same animation. GIF is older and simpler: at most 256 colours per frame, and a pixel is either fully transparent or fully opaque. That is why a GIF made from a lossy WebP is usually bigger, and why a smooth gradient can show banding after the conversion. You convert anyway because many places still accept GIF but not WebP, including some older email clients, chat apps, and design tools. The GIF will not match the WebP bit for bit. What you can soften, before you convert, is width, the matte on soft edges, and the colour budget.</p>
<h2>The frames are already there; they are just stored differently</h2>
<p>If the WebP moves in the before pane, it is an animated file and the conversion should produce an animated GIF. If it does not move, it is a still image and the result is a one-frame GIF, which is still useful when a form insists on that format. The converter is not inventing motion and it is not joining still photographs. It is reading the file you dropped.</p>
<p>A WebP animation does not always store every frame as a complete picture. Each frame can be a rectangle smaller than the whole canvas, placed at an offset, with rules for blending and for clearing what was there before. If those partial rectangles were copied into a GIF as they stand, playback would be wrong. The converter applies those rules and rebuilds every full frame before the GIF is written, so what you see in the result is what the WebP shows.</p>
<p>Timing is kept, with the limits of the destination format. Each frame keeps its own delay from the WebP, rounded to the hundredths of a second that GIF can store, and never shorter than two hundredths of a second, because browsers raise lower values. Delays of 0 or under 20 milliseconds are treated the way browsers treat them and become 100 milliseconds. That last rule matters if the WebP has a very fast burst: a handful of nearly instant frames will not stay nearly instant in the GIF. They will each hold for a tenth of a second.</p>
<p>One more loop detail is easy to miss. The loop count is not carried over. The GIF always loops forever. If the original WebP was meant to play once and stop, the GIF will not preserve that. You get the frames and their delays, not the original repeat policy.</p>
<h2>Soft edges need a matte, because GIF has no partial transparency</h2>
<p>Transparency is the other place the formats disagree. The conversion can keep transparency: transparent pixels stay transparent. Soft edges are a different problem. GIF has no partial transparency, so pixels that are mostly see-through become fully transparent and the rest become opaque. The soft edge in between is blended with a matte colour you choose, so the fringe does not look jagged against a page of a similar colour. Pick the matte to match where the GIF will sit. If you would rather not keep a see-through background at all, you can switch transparency off and fill the background with a colour.</p>
<p>That choice decides whether a sticker or an emoji still looks cut out after the conversion, or whether a halo appears once the file leaves a white preview and lands on a dark message bubble. The page’s own example of a lossless animated WebP with an alpha channel keeps the transparent background and blends the edge pixels with a white matte. At 320 × 320 pixels and 16 frames, that WebP is 296 KB and the GIF is 356 KB, with transparency kept and a balanced colour budget. A white matte is the right match for a white page, and the wrong one for a black one.</p>
<h2>The GIF is usually larger, and width is the first lever</h2>
<p>People often notice the file size first and assume something went wrong. WebP compresses animations far better than GIF can, so a larger GIF is the expected outcome. The page shows this with its own samples.</p>
<p>A lossy 12-frame neon Open sign at 480 × 270 pixels goes from 32 KB as WebP to 163 KB as a GIF, with the frame count and each frame’s delay kept. A 15-frame bounce on a flat coloured background, also 480 × 270, goes from 24 KB to 123 KB: simple flat colour helps, but does not close the gap. A mixed lossy and lossless 16-frame moon, converted at 240 pixels wide instead of the original 320, still grows: 147 KB at 320 × 320 becomes a 305 KB GIF at 240 × 240. The width cut helps, and the format change still wins on size.</p>
<p>The page calls a smaller width the quickest way to shrink the GIF. You can keep the original width or pick a smaller one. The output is limited to 1024 pixels on the longer side, so a very wide or tall WebP cannot pass through at full size. After width, File size is Balanced or Smaller file, which sets how many colours the GIF uses. Fewer colours can reduce the weight and can make banding in gradients more obvious. If the file is still too large afterwards, a compressor or a resizer on the same site can shrink it further, including toward a target such as 256 KB, or to an exact width and height. That is a later step, not part of the conversion itself.</p>
<p>There are hard caps as well. The drop box accepts a .webp file up to 25 MB and up to 300 frames. The GIF is capped at 300 frames and 30 seconds of playing time. A long or very dense animation may not come through in full.</p>
<h2>How the conversion runs, and what it will not open</h2>
<p>The practical path is short, and I use <a href="https://loopgif.net/webp-to-gif">LoopGIF’s WebP to GIF tool</a> for it. You drop a .webp file onto the box or click to choose one. Animated and still files both work. The left pane plays the WebP as your browser shows it, with its size and frame count, and a line under the drop box says whether it has transparency. You choose a width and what to do about transparency, then press Convert to GIF. The animated GIF appears beside the original with its size, and Download saves it.</p>
<p>All of that happens in the browser. The file is read and converted on your device. Nothing is uploaded, no account is needed, and there is no watermark. The converter is free and does not require a sign-up. Animated files are read by the browser’s own decoder where it supports WebP animation, and by the page’s own WebP reader otherwise; either way, each frame keeps its delay. That fallback is the reason a browser that can display a still WebP but not an animated one may still be able to convert the animation.</p>
<p>If a file will not open, the usual causes are specific. The file may not be a valid WebP, it may be over 25 MB or 300 frames, or the browser may not support WebP. The message under the drop box says which. A WebP that the browser cannot decode at all cannot be opened, and a notice appears when the browser lacks WebP support. The tool was built and tested in current Chrome. The page also checks the browser you are using, by asking it directly rather than guessing from the name, for WebP images and for background encoding with Web Workers.</p>
<p>This is also the place to be clear about what the tool is not. It converts an existing WebP, animated or still, into a GIF. It does not convert a GIF back into a WebP. It does not assemble a folder of still photographs into a new loop, and it does not trim or transcode a video. If you need a new animation rather than a translation of one you already have, that is a different job.</p>
<h2>When this translation is the right job</h2>
<p>The cases are ordinary, which is why the format gap keeps showing up. Some forms, forums, and editors accept GIF but not WebP. GIF still animates in the widest range of mail and chat apps, which is often why a short loop should be converted before it is sent. A transparent animated WebP can become a sticker or an emoji with the background still see-through, if you keep transparency and choose a matte that matches the place it will sit. An animated WebP you saved from the web will not play in every app; a GIF will. And a still WebP can become a one-frame GIF when a form insists on that format and you do not want to invent motion you do not have.</p>
<p>In each of those cases, the quality of the result depends on the same three decisions: whether you keep the original width or give some of it away to shrink the file, whether you keep transparency and with which matte, and whether you ask for a Balanced colour budget or a Smaller file. The timing of each frame comes across, within the hundredths GIF can store. The loop will run forever. The GIF will almost certainly be larger than the WebP you started with. If you go in knowing those tradeoffs, the conversion does what it is for: it takes an animation you already have and makes a file more places can actually play.</p>
<p><em>I build LoopGIF.</em></p>
]]></content:encoded></item><item><title><![CDATA[A Set of Stills Needs an Order, a Hold on Each Picture, and One Shared Frame]]></title><description><![CDATA[You already have the stills. Five angles of a product, a before and an after, a stack of tutorial screenshots, or a weekend of photos you would rather send…]]></description><link>https://loopgif.hashnode.dev/a-set-of-stills-needs-an-order-a-hold-on-each-picture-and-one-shared-frame</link><guid isPermaLink="true">https://loopgif.hashnode.dev/a-set-of-stills-needs-an-order-a-hold-on-each-picture-and-one-shared-frame</guid><category><![CDATA[gif]]></category><category><![CDATA[animation]]></category><category><![CDATA[Design]]></category><category><![CDATA[Tutorial]]></category><category><![CDATA[webdev]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 13:52:50 GMT</pubDate><content:encoded><![CDATA[<p>You already have the stills. Five angles of a product, a before and an after, a stack of tutorial screenshots, or a weekend of photos you would rather send as one file than as ten. What you want is a loop: picture, picture, picture, then back to the first. The failure is rarely that the pictures cannot be joined. It is that they are joined in the wrong order, held for the same time when one of them has to be read, or forced into a frame that crops the useful edge or leaves bars you did not expect.</p>
<p>Assembling stills into a GIF is not the same as taking an existing GIF apart into numbered pictures, and it is not the same as asking a single photo to move on its own. This is a flipbook. The pictures you already have become frames. They play in the order you set, each for the time you give it, and they all have to share one frame size.</p>
<p><em>I build LoopGIF.</em></p>
<h2>The list order is the play order</h2>
<p>Add the pictures in the sequence you want, or add them in any order and move them afterwards. You can drop them onto the box, choose them from the file picker, or paste them from the clipboard. They appear as thumbnails in order. Drag those thumbnails, or use the arrow buttons if you are on a phone and dragging is awkward. The first thumbnail is the first frame of the loop. If a closing product shot lands in slot one, the loop will start there.</p>
<p><a href="https://loopgif.net/image-to-gif">LoopGIF’s image to GIF tool</a> is built for this assembly job: several images in, one animated GIF out. The first picture also sets the aspect ratio of the GIF, so it decides both the opening frame and the canvas the rest of the set will be fitted into.</p>
<p>The converter accepts up to 60 images, each up to 15 MB, as JPEG, PNG, WebP, or GIF. If you drop an animated GIF, only its first frame is used. This tool does not take the rest of that animation apart, and it does not change the speed of an existing animated GIF. HEIC files are not accepted, so export them as JPEG first. Once the pictures are in the list, you can also duplicate or remove frames, which helps when a still should appear twice, or when an extra shot made it into the set by mistake.</p>
<h2>Each still has its own hold</h2>
<p>Each picture has its own delay, from 20 milliseconds to 10 seconds. A second is 1000 ms. The delay is stored in whole hundredths of a second, and it is never stored below 2 hundredths, because browsers play anything shorter at about 10. That is why the smallest delay the page offers is 20 ms.</p>
<p>Give a long delay to the picture people need to read, and a short one to a quick flip. The page’s own suggested ranges are specific. Four or five sides of a product work at 600 to 800 ms and then loop. A before and after is clearer at about 1.5 seconds on each picture. Tutorial screenshots can be held for two or three seconds so a step can actually be read. A recap of 10 to 20 photos sits well at about 800 ms. Hand-drawn frames can be set at 100 ms, or at 40 to 80 ms if you want smoother motion.</p>
<p>You do not have to type every field by hand. Set one value in Delay for every frame and press Apply to all, then change single frames afterwards. A uniform delay is the usual mistake: ten product photos at 200 ms will flicker past before anyone can see the label, and a recap held for three seconds a picture feels like a slideshow that never ends. The delay belongs to the still, not to a global frame rate you picked once and hoped would fit every picture.</p>
<h2>One GIF means one frame size</h2>
<p>Still images almost never arrive at the same size. A square product shot, a wide landscape from the same weekend, and a tall phone screenshot cannot sit in one GIF until they share one frame. Frames in a single GIF have to be the same size. The first picture sets that shape. The others are fitted into it.</p>
<p>There are three ways to fit. Contain shows the whole picture and fills the leftover space with a background colour you choose; nothing is cropped. Cover fills the frame and crops the edges. Stretch distorts the picture to fit, and the page suggests stretch only when the screenshots already share a shape.</p>
<p>The output width can be 320, 480, or 640 pixels, or the original, and never more than 800 pixels on either side. Pictures are downscaled when they are loaded, to keep memory use low.</p>
<p>The page shows what Contain does with mismatched shapes: three pictures of different shapes, fitted inside a 480 by 270 frame so nothing is cropped, each held for 1000 ms, with no crossfade, contain on #17141f, 480 px wide, looping forever, in the smaller file mode. The result is listed as 3 frames, 30 KB, and a 3 second loop. That 480 by 270 frame is the shape the first picture set, at a 480-pixel width. For a product flipbook with nothing cropped, the page suggests contain and a white background. For a recap where every photo fills the frame, it suggests cover.</p>
<h2>A cut and a crossfade are different file costs</h2>
<p>A hard cut from one still to the next is the default. Crossfade is optional, and it is the setting that changes the file the most. It inserts blended frames between neighbouring pictures, from 1 to 8 of them, each lasting 0.05 seconds. Every one of those frames is a new full picture, so a three-frame crossfade between two pictures makes the GIF several times larger than a plain cut. The page says crossfade is the biggest cost.</p>
<p>The output can hold at most 300 frames in all, counting those blended frames. The slider lowers itself when you add many pictures. With 60 images, the crossfade can reach 4 frames, and the limit is applied automatically. There is also an option to fade the last image back to the first, so the loop does not jump when it repeats.</p>
<p>The page’s crossfade example is two pictures, each held for 900 ms, with a three-frame crossfade and the last image fading into the first, cover, 320 px wide, looping forever, in the smaller file mode. The result is listed as 8 frames, 194 KB, and a 2.1 second loop. Two source pictures became eight frames because the blends, including the wrap back to the first picture, were added in. A four-picture flipbook at 500 ms each, with no crossfade, cover, and 320 px, is listed as 4 frames, 172 KB, and a 2 second loop.</p>
<p>Photos are heavy as GIFs, because every frame is a full picture with at most 256 colours. The page shows the frame count and an estimated size before you export. The estimate comes from encoding a few sample frames and is usually within about 25 percent of the real file. To make the file smaller, pick a narrower width, use fewer pictures, turn the crossfade down, or choose the smaller file mode, which uses 127 colours and ignores tiny colour changes between frames. It usually gives a lighter file with a little less colour accuracy.</p>
<h2>What the tool writes, and where it stops</h2>
<p>The steps on the page are short. Add your images. Order and time them. Choose the look: a crossfade or not, how pictures fit the frame, a width of 320, 480, 640, or the original, and a loop of forever, once, or a number of plays (2 to 100). Check the live preview, the frame count, and the estimated file size, then press Download. The GIF is encoded in your browser in a few seconds.</p>
<p>The converter is free, needs no sign-up and no account, and runs in your browser. Your images never leave your device. They are decoded, drawn on a canvas for the live preview, and encoded when you press the button. There is no upload. No watermark is added. The page says it works in current Chrome, Edge, Firefox, and Safari.</p>
<p>Transparent output is available when at least one picture has transparent pixels, such as a PNG cut-out or a WebP sticker. The empty space around contained pictures then stays see-through too. GIF transparency is on or off, with no soft edges, so edge pixels are blended toward an edge colour that you choose. Pick the colour of the page the GIF will sit on, to avoid a halo.</p>
<p>This converter stays free. Making a single photo move, rather than flipping through a set of photos you already have, is a different tool: an AI photo animator that needs an account and costs 1 credit per GIF. One still that is asked to move is not a set of stills assembled in order.</p>
<h2>Check the order, the holds, the fit, and the loop</h2>
<p>Before you download, look at four things. First, the order: the first thumbnail is first, and it also sets the shape. If the closing shot is in slot one, the loop starts there, and the rest of the set is fitted to that shot’s aspect ratio. Second, the holds: a step people need to read should not get the same 500 ms as a flipbook frame. Watch the live preview once through. Third, the fit: if a source has a different shape from the frame the first picture set, decide whether you are looking at bars, at cropped edges, or at a stretched picture, and whether the background colour is the one you meant. Fourth, the loop and the size: forever, once, or a set number of plays are different files. If you used a crossfade, decide whether the last image should fade back to the first. If the estimate is too heavy, narrow the width, drop a picture, lower the crossfade, or switch to the smaller file mode, and watch the estimate update.</p>
<p>If those four look right, download the new file. A GIF made from stills is not a shorter version of an existing loop, and it is not a single photograph asked to move. It is one shared frame, a play order, and a hold that belongs to each picture.</p>
<p><em>I build LoopGIF.</em></p>
]]></content:encoded></item><item><title><![CDATA[A One-Way Reverse Still Jumps at the Seam; a Boomerang Travels Back]]></title><description><![CDATA[You already have a GIF that travels one way: a wave, a blink, a product turning to show its other side. On repeat, the last frame jumps back to the first…]]></description><link>https://loopgif.hashnode.dev/a-one-way-reverse-still-jumps-at-the-seam-a-boomerang-travels-back</link><guid isPermaLink="true">https://loopgif.hashnode.dev/a-one-way-reverse-still-jumps-at-the-seam-a-boomerang-travels-back</guid><category><![CDATA[gif]]></category><category><![CDATA[animation]]></category><category><![CDATA[Design]]></category><category><![CDATA[Tutorial]]></category><category><![CDATA[webdev]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 13:47:08 GMT</pubDate><content:encoded><![CDATA[<p>You already have a GIF that travels one way: a wave, a blink, a product turning to show its other side. On repeat, the last frame jumps back to the first, and the cut is obvious. The usual fix people reach for is to reverse the file. That only plays the motion backwards. When the reversed file loops, it still has to jump, from its new last frame back to its new first frame, which is the old last frame. The path never returns. A boomerang is a different construction. It plays forward and then backward, so the motion comes back along the same frames and meets itself without that snap.</p>
<p><em>I build LoopGIF.</em></p>
<p>The two are easy to confuse because both change direction. They are not the same loop. A reversed GIF plays only backward, once through, and then starts that backward pass again. A boomerang is a palindrome: out, then back, again and again. It is built to return.</p>
<h2>Reverse changes direction. It does not close the path.</h2>
<p>A reversed GIF is a one-way trip in the other direction. The frames run from last to first. When the file repeats, it jumps back to its new first frame, which is the old last frame. If the original was a one-way motion, the repeat looks abrupt: the object finishes travelling backward and is suddenly at the end of the original path again. The jump is not a defect in the reverse step. It is what “play backwards, then start over” means.</p>
<p>A boomerang does not start over at the far end. It turns around. The frames play in order, then in reverse order, and only then does the file meet the first frame again, because the return trip has already brought you there. A fall can rise along the same path. A wave can rock. A product can turn one way and back. The seam is a turn, not a cut from last to first.</p>
<p>A boomerang is not automatically the right loop for every file. It works best when the motion makes sense in both directions: a wave, a bounce, a blink, a spinning object. A person walking forward and then backward looks odd. A GIF that already loops smoothly is often better left as it is, because the boomerang doubles the frames and the file grows. If you only wanted a rewind that does not have to sit on a page forever, a one-way reverse may still be the right file. If you wanted a loop that can repeat without jumping from the last frame to the first, you need the return.</p>
<h2>The turns stutter if the end frames appear twice</h2>
<p>Building the return naively is a simple ping-pong: play the frames in order, then play them in reverse order. If the reverse run starts with the frame you just showed, that frame appears twice in a row. The same thing happens at the start of the loop, when the return arrives at the first frame and the next cycle starts on it again. The motion seems to catch for a moment at each turn. That is the stutter.</p>
<p><a href="https://loopgif.net/boomerang-gif">LoopGIF’s boomerang GIF tool</a> avoids that hitch by default. A control labelled Skip the repeated end frames shows the first and last frame once at each turn. A GIF with N frames becomes a GIF with 2N minus 2 frames: the first and last frame appear once, and the frames in between appear twice, once on the way out and once on the way back. You can turn the option off to see the difference, or if you want the end frames to linger.</p>
<p>The live page publishes examples that make the count visible. I did not re-export them; these are the page’s figures. A wave that already loops becomes a rocking motion. With the end frames shown once, the 20 frames become 38. The settings are forward then back, end frames shown once, hold 0 ms at both ends, and speed 1 times. The file is listed as 91 KB at 480 × 270 becoming 172 KB at the same size.</p>
<p>With Skip the repeated end frames turned off, a 12-frame GIF becomes 24, and the first and last picture each stay on screen for two frames at the turns. The page says this is the stutter the default setting removes. The settings are forward then back, end frames repeated, hold 0 ms at both ends, and speed 1 times. The file is listed as 46 KB at 480 × 270 becoming 83 KB at the same size. The arithmetic matches the rule: skipped ends give 2N minus 2; kept ends give 2N. Before you make the file, the page shows the frame count and the length of the new GIF.</p>
<h2>A pause is time added at the turn</h2>
<p>Returning is not the same as pausing, and pausing is not a rewrite of every delay. Every frame keeps its own delay, forward and backward, so a GIF with uneven timing stays uneven. A hold that already belonged to a frame still belongs to it on the way out and on the way back.</p>
<p>If you want the file to rest at a turn, that is a separate control. A pause adds time to the frame at the turn: the pause at the far end is added to the last frame of the GIF, and the pause at the start is added to the first frame. Each pause can be up to 2 seconds. The far end is the last frame of the original.</p>
<p>The page’s default-style example is a cat that holds its thumbs up. The default boomerang with a 500 ms hold at the far end turns 16 frames into 30. The cat holds the pose for half a second, and then it goes back. The settings are forward then back, end frames shown once, hold 500 ms at the far end and 0 ms at the start, and speed 1 times. The file is listed as 337 KB at 320 × 320 becoming 638 KB at the same size, with 30 frames in all.</p>
<p>You can also change the speed, from a quarter to four times. The speed control scales all the delays. It is a multiplier on the boomerang you are building, not a substitute for the return itself, and it is not a way to set one fixed delay for every frame. Delays are whole hundredths of a second and never below 2, so a very fast boomerang of a GIF that is already fast cannot go faster than that limit.</p>
<p>One published example combines both: speed 1.5 times, with a 300 ms pause at the start and at the far end. The 12 frames become 22, and the whole boomerang lasts about 1.4 seconds. The settings are forward then back, end frames shown once, hold 300 ms at both ends, and speed 1.5 times. The file is listed as 130 KB at 480 × 270 becoming 237 KB at the same size, with 22 frames in all. Twelve frames with the ends skipped would be 22, which matches 2N minus 2. The pauses change how long those end frames are shown; they do not add extra frames.</p>
<h2>What the tool accepts, and where it stops</h2>
<p>Drop a GIF onto the box or click to choose one. Leave Skip the repeated end frames on for a smooth turn, or turn it off to show the first and last frame twice. Add a pause at the far end or at the start if you want one, and change the speed if you want to. The preview plays the result. Press Make boomerang. The new GIF appears next to the preview with its frame count, size and dimensions, and Download saves it.</p>
<p>The tool accepts a GIF only, up to 25 MB and up to 300 frames. It is free, needs no sign-up and no account, and runs in your browser. Your GIF never leaves your device: it is opened, processed and written in the browser, with no upload. The page’s structured data also lists no watermark among the features. The result is capped at 300 frames, 30 seconds, and 1024 px on the longer side. Because the boomerang has about twice as many frames as the original, the file is larger, and a long GIF can reach the 300 frame limit. The page tells you before you start.</p>
<p>The boomerang plays forward first, and then backward. There is no option to start backward. If you need the first move to be a rewind, you would have to reverse the GIF first, in a separate step, and then build the boomerang from that reversed file. This page is for the boomerang itself, with pauses and clean turns. Transparent GIFs stay transparent, and soft edges are blended with white.</p>
<p>The tool was built and tested in current desktop Chrome, using ordinary canvas, workers, and GIF decoding. The page has not tested other browsers itself. A live “Does your browser support this?” list is checked against the browser you are using by asking it directly, not guessed from the browser name. The checks named on the page are GIF files and background encoding (Web Workers).</p>
<h2>Watch the return, then watch the turns</h2>
<p>Before you download, look at three things. First, whether the motion should return at all. Play the preview all the way out and all the way back. A wave, a bounce, a blink, or a spinning object will usually read as a loop. A one-way action that looks wrong in reverse will still look wrong, only now twice per cycle.</p>
<p>Second, the turns. With Skip the repeated end frames left on, the first and last frame should appear once, and the motion should not catch at either end. If you turned the option off, expect those frames to linger. Then check any pause you added: the far-end pause should sit on the last frame of the original, and the start pause should sit on the first.</p>
<p>Third, the numbers. The new file will have about twice the frames, or exactly 2N minus 2 with the ends skipped, and it will usually be larger. If the source was already long, confirm that the result stays under 300 frames and 30 seconds. The page shows the frame count and the length before you make the file.</p>
<p>If those three look right, download the new GIF. A boomerang is not a reversed file with a hope that the seam will soften. It is a palindrome: forward, then back, with the end frames handled on purpose so the turns do not hitch. The loop closes because the motion returns, not because the file jumps to the start.</p>
<p><em>I build LoopGIF.</em></p>
]]></content:encoded></item><item><title><![CDATA[Splitting a GIF Exports Numbered Stills, Not a Shorter Loop]]></title><description><![CDATA[You already have the animation. What you need now is not another GIF. You need the pictures inside it: a poster still, a run of frames to retouch, or…]]></description><link>https://loopgif.hashnode.dev/splitting-a-gif-exports-numbered-stills-not-a-shorter-loop</link><guid isPermaLink="true">https://loopgif.hashnode.dev/splitting-a-gif-exports-numbered-stills-not-a-shorter-loop</guid><category><![CDATA[gif]]></category><category><![CDATA[animation]]></category><category><![CDATA[Design]]></category><category><![CDATA[Tutorial]]></category><category><![CDATA[webdev]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 13:40:32 GMT</pubDate><content:encoded><![CDATA[<p>You already have the animation. What you need now is not another GIF. You need the pictures inside it: a poster still, a run of frames to retouch, or numbered images for a game engine that wants frames rather than a loop. A trim and a crop both miss that job. A cut keeps a run of frames and writes a shorter GIF, so you still have an animation, only briefer. A crop keeps a rectangle of every frame, so the picture gets smaller and the length stays the same. Neither gives you a folder of stills.</p>
<p><em>I build LoopGIF.</em></p>
<p>The work is to take the file apart and look at every frame at once. Load the GIF, wait while the thumbnails are drawn, and you get a grid. Each cell shows the frame, its number, and its delay in milliseconds, which tells you how long that frame is on screen. Tick the frames you want and download them as a ZIP of PNG or JPG files. The original loop is not rewritten. What you take away is stills.</p>
<h2>A still is not a cut, and it is not a crop</h2>
<p>It is easy to treat “get something out of this GIF” as one problem. It is at least three. Cutting is a time-range job: you keep some frames, drop the rest, and the result is still a GIF. Cropping is a spatial job: you keep part of the picture on every frame, and the result is still a GIF, just smaller in width and height. Splitting is a different deliverable. The output is one image per selected frame, packed into a ZIP, named so you can tell which frame each file came from.</p>
<p>That difference matters when you are editing frame by frame, hunting for a poster, or packing assets for a game. You do not want a shorter loop. You want pictures you can open, retouch, and, if you need a lighter animation later, rebuild in another tool. This splitter does not cut a time range, does not crop the frame, and does not join, reverse, or retime the GIF. It takes the animation apart.</p>
<p>If you only need one frame, a tool that lets you scrub to a single frame and save it as a PNG is enough. The splitter is for the other case: seeing all the frames in a grid and downloading many of them at once.</p>
<h2>Each file is the complete picture, not the patch the GIF stored</h2>
<p>A GIF often stores only the part of the picture that changed. If you lifted those stored regions out as files, you would not get a usable still. You would get the update, not the frame. <a href="https://loopgif.net/gif-splitter">LoopGIF’s GIF splitter</a> draws each selected frame at its full size, as a browser would draw it at that moment, with earlier frames composited in, not just the part that changed. Each file in the ZIP is that complete picture. That is what you want for editing or for a game.</p>
<p>The grid is only a map of those frames, not the export itself. The cells are thumbnails of up to 160 px. The files in the ZIP are full size. A line under the grid counts the thumbnails off as they are drawn, so you can see the file being read before you choose anything. The cells are check boxes, so you can click them or tab to them and press Space.</p>
<p>The delay on each cell is there to be read. If you are studying the animation, or you plan to rebuild it later, the milliseconds tell you how long that frame is held on screen. The ZIP itself is a set of stills.</p>
<h2>You rarely want every frame, and the names should still say which ones you took</h2>
<p>Every frame starts ticked, which is the right default if you want the whole stack. You rarely want that. All and None are the two ends. Every 2nd picks frames 1, 3, 5 and so on, which halves the ZIP and is a fast way to thin out a long animation. Invert flips the selection, so you can untick a few frames you do not want and then invert to get exactly those. You can also tick and untick single frames, or select a run when you only need part of the animation. The count of selected frames is in the heading above the grid.</p>
<p>The selected frames keep their original numbers in the file names. The files are named like your-gif-frame-001.png, using each frame’s own number, padded to three digits, or more if the GIF has over 999 frames. If you tick frames 4 to 9, the files are called name-frame-004 up to name-frame-009. They still sort in order, and you can match them back to the GIF. Every 2nd leaves the numbers skipped, because the names are the frame’s original position in the GIF, not a new sequence from 001.</p>
<p>Press Download and the tool encodes each selected frame and puts it in a ZIP. The ZIP is made on your device and handed to your browser as a download, with a name like your-gif-frames-png.zip. The ZIP stores the images without extra compression, because PNG and JPG are already compressed, so building it is quick.</p>
<h2>PNG keeps the pixels. JPG asks you to fill the transparent ones.</h2>
<p>Choose the format before you download. PNG keeps transparency, is lossless, and keeps every pixel exactly. Use it when the still has to stay faithful, or when a transparent background has to survive.</p>
<p>JPG is smaller and needs a background colour and a quality setting. JPG cannot hold transparency, so every transparent pixel becomes the background colour you choose. If you pick JPG, you are also picking a fill.</p>
<p>The page publishes four examples, each splitting one of its own GIFs the way the tool does, with a contact sheet of the ZIP files and their real names. I did not re-export them; these are the page’s figures.</p>
<p>The first takes every frame of a 16 frame GIF as PNG files, one file per frame, numbered from 001. The settings are frames all 16, saved as PNG files in a ZIP. The files are listed as a 34 KB GIF at 480 × 270 px becoming a ZIP of 16 PNG files, 77 KB.</p>
<p>The second takes every 2nd frame of a 33 frame typing animation, saved as JPG, with the original frame numbers kept so the skipped numbers still identify the frames. The settings are frames every 2nd (1, 3, 5 and so on), saved as JPG files in a ZIP. The files are listed as a 28 KB GIF at 480 × 270 px becoming a ZIP of 17 JPG files, 108 KB.</p>
<p>The third takes frames 1 to 12 of a spinning logo as PNG files, a run for when you only need the first part of an animation. The settings are frames 1 to 12, saved as PNG files in a ZIP. The files are listed as a 123 KB GIF at 320 × 320 px becoming a ZIP of 12 PNG files, 201 KB. Those twelve files are stills, not a trimmed logo GIF.</p>
<p>The fourth takes frames 5 to 12 saved as JPG at quality 80. JPG files of flat colour are small, so this ZIP is one of the smallest. The settings are frames 5 to 12, saved as JPG files in a ZIP. The files are listed as a 91 KB GIF at 480 × 270 px becoming a ZIP of 8 JPG files, 72 KB.</p>
<h2>What the tool accepts, and where it stops</h2>
<p>The steps on the page are short. Drop a GIF onto the box or click to choose one. Up to 25 MB and 300 frames are accepted, and the file has to be a GIF. All frames appear in the grid as thumbnails. Select the frames, pick PNG or JPG, and press Download. The tool saves each selected frame at full size and zips them.</p>
<p>The tool is free, needs no sign-up and no account, and runs in your browser. Your GIF never leaves your device. The frames are drawn, encoded and zipped in the browser, with no upload. The download is created on your own device. The page’s structured data also lists no watermark among the features.</p>
<p>The ZIP is built in memory. A very large GIF with many big frames can take a while and a lot of memory, so there is a progress bar and a Cancel button.</p>
<p>The page says the tool was built and tested in current Chrome. The ZIP is a standard file that opens in every operating system. There is a live “Does your browser support this?” list, checked against the browser you are using right now by asking it directly, not guessed from the browser name. The check named on the page is GIF files.</p>
<p>The usual reasons are practical: retouch frames and rebuild the animation in another tool, pick the sharpest or funniest still as a poster, pack numbered images for an engine that wants frames, study the timing on each cell, or select every second frame and rebuild a lighter animation later. In each case, the deliverable is the stills.</p>
<h2>Check the selection, the format, and the numbers on the files</h2>
<p>Before you download, look at three things.</p>
<p>First, the selection. The count in the heading is the number of files you are about to get. If you only need a run, tick that run. If you are thinning a long GIF, Every 2nd is the fast path. If you want a handful of frames and it is easier to turn the rest off than to turn those on, use Invert.</p>
<p>Second, the format. PNG if the still has to keep transparency, or if every pixel has to stay exact. JPG if a smaller file is worth choosing a background colour and a quality setting, and you are willing to fill every transparent pixel with that colour.</p>
<p>Third, the names. The numbers on the files are the original positions in the GIF. They will skip if you skipped frames, and that is useful, not a bug. You should still be able to match a file back to its cell in the grid.</p>
<p>If those three look right, press Download. What you get is a ZIP of stills, each one the complete picture as it would have been drawn at that moment. It is not a shorter GIF, and it is not a cropped one. It is the frames, taken out and named so you can still tell which is which.</p>
<p><em>I build LoopGIF.</em></p>
]]></content:encoded></item><item><title><![CDATA[Joined GIFs Keep Their Own Timing, but They Have to Share One Frame Size]]></title><description><![CDATA[You already have the clips, and they are already GIFs. What you want is one loop: the first file all the way through, then the second, then maybe a third…]]></description><link>https://loopgif.hashnode.dev/joined-gifs-keep-their-own-timing-but-they-have-to-share-one-frame-size</link><guid isPermaLink="true">https://loopgif.hashnode.dev/joined-gifs-keep-their-own-timing-but-they-have-to-share-one-frame-size</guid><category><![CDATA[gif]]></category><category><![CDATA[Design]]></category><category><![CDATA[animation]]></category><category><![CDATA[Tutorial]]></category><category><![CDATA[webdev]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 13:32:28 GMT</pubDate><content:encoded><![CDATA[<p>You already have the clips, and they are already GIFs. What you want is one loop: the first file all the way through, then the second, then maybe a third, and then the whole sequence again. Two assumptions usually sneak in, and both are wrong. The join does not average the timing. A slow GIF stays slow, and a fast one stays fast, because each frame keeps its own delay. And the join cannot leave each file at its own width and height: every frame in a single GIF has to be the same size, so mismatched sources have to be fitted into one shared frame.</p>
<p><em>I build LoopGIF.</em></p>
<p>The result is sequential, not simultaneous. All of the first GIF plays, then all of the second, and so on, and then the whole thing loops. The second starts as soon as the first ends. If the sources differ in shape, each one is fitted into that one frame: scaled down to sit inside it, with bars of a background colour where the shape differs, or scaled up until it fills the frame, with whatever sticks out trimmed away. Transparency does not survive that step, because the background colour is painted behind every frame.</p>
<h2>The list order is the play order</h2>
<p>Add the GIFs in the order you want to see them, or add them in any order and move them afterwards. The list order is the play order. The first row is first. <a href="https://loopgif.net/combine-gifs">LoopGIF’s combine GIFs tool</a> uses up and down arrow buttons on each row to move a GIF, and an X to remove one. The page is explicit about using the arrows rather than dragging, which also works with a keyboard and on a touch screen. The names in the list are your file names, so you can tell the rows apart.</p>
<p>This is an end-to-end join, not a grid. The tool plays the files one after another. If you wanted them to play at the same time, side by side, this tool does not do that. After the last GIF finishes, the combined file does not stop. The whole sequence loops, so the last frame of the last GIF is followed by the first frame of the first.</p>
<p>The page publishes a same-shape example that shows the basic sequence. I did not re-export it; these are the page’s figures. Two GIFs of the same shape are joined at the first GIF’s size. The second starts as soon as the first ends, and each keeps its own frame delays. The settings are listed as 2 GIFs, output the first GIF’s size (480 × 270 px), fit contain on #ffffff. The files are listed as 164 KB at 480 × 270 becoming a 160 KB GIF at the same size, with 28 frames.</p>
<h2>Delays survive the join. They are not averaged.</h2>
<p>Timing is per frame. The tool does not stretch or shrink a GIF to match another. A GIF with 10 frames at 100 ms and another with 40 frames at 20 ms keep their own speeds. A slow GIF stays slow and a fast one stays fast. Every frame keeps its own delay, so each GIF plays at its own speed in the combined one.</p>
<p>That is usually what you want: a held title should still hold, and a quick reaction should still be quick. The surprise comes when you expected the join to smooth them into one tempo. This tool does not retime GIFs to match each other. If you need the speeds to match, you would have to change one of the files first, in a separate step.</p>
<p>The total length is a sum, not an average. The page’s three-GIF example states this directly: three GIFs of different lengths joined at a width of 320 px. The total frame count is the sum of the three, and the total time is the sum of the three times. The settings are listed as 3 GIFs, output 320 px wide, fit contain on #fff8ef. The files are listed as 253 KB at 480 × 270 becoming a 152 KB GIF at 320 × 180, with 44 frames. The height follows the first GIF’s shape.</p>
<p>Because the output is a sum, the limits are on the combined file, not only on each source. You can add up to 12 GIFs, each up to 25 MB, but the combined GIF is capped at 300 frames and 30 seconds of playing time. The summary line shows the total frames, the total time, and the output size before you start. If you go over, the page shows the numbers and does not start.</p>
<h2>One GIF means one frame size</h2>
<p>GIF frames are all the same size, so when your GIFs are not, each has to be fitted into the output size. A square reaction and a wide screen recording can look fine on their own, and still fail to share a single canvas until they are fitted.</p>
<p>The output size is the first GIF’s size by default, so the first GIF is never scaled. That is why the first row matters for more than play order: it also sets the default frame. If you would rather have a standard width, choose 320, 480, or 640 pixels wide, and the height follows the first GIF’s shape. The output is limited to 1024 px on the longer side.</p>
<p>Contain and Cover are the two ways to handle a different shape. Contain scales the GIF down to fit inside, keeping its shape, and fills the leftover space with the background colour you pick. The whole GIF remains visible. Cover scales it up until it fills the frame, and trims whatever sticks out. There are no bars, but the edges are gone. The background colour defaults to #FFFFFF on the page, and you can change it. Whatever you pick is painted behind every frame, so the combined GIF is not transparent. If one of the sources had a transparent background, that transparency is not kept.</p>
<p>The page’s Cover example is a square GIF and a wide one joined into a square. The wide one fills the frame and its edges are cropped, so there are no bars. The settings are listed as 2 GIFs, output 240 px wide, fit cover on #17141f. The files are listed as 411 KB at 320 × 320 becoming a 290 KB GIF at 240 × 240, with 36 frames.</p>
<p>The Contain example that makes the bars obvious is a 360 by 240 GIF and a 468 by 60 banner. With Contain, the banner is scaled to fit the width and sits in the middle of the frame with black bars above and below. The settings are listed as 2 GIFs, output the first GIF’s size (360 × 240 px), fit contain on #000000. The files are listed as 211 KB at 360 × 240 becoming a 209 KB GIF at the same size, with 55 frames. The first GIF, at the default output size, is not scaled. The banner is the one that has to be fitted.</p>
<p>Choose the mode by what you cannot afford to lose: Contain keeps the whole GIF visible, at the cost of bars; Cover fills the frame, at the cost of trimmed edges.</p>
<h2>What the tool writes, and where it stops</h2>
<p>The steps on the page are short. Drop two or more GIFs, or click to choose them. They appear in a list with their size, frames, and length. Put them in order with the arrows. Choose the output size and the fit. Press Combine GIFs. The new GIF appears with its size, frames, and length, and Download saves it.</p>
<p>The tool accepts GIF files only. It is free, needs no sign-up and no account, and runs in your browser. Your GIFs never leave your device. They are opened and joined in the browser, with no upload. Nothing is sent to a server. The page’s structured data also lists no watermark among the features.</p>
<p>The combined GIF is a fresh encode with one shared palette of up to 256 colours, so a mix of very different GIFs may show some colour banding. That is worth a look when you join a flat graphic to a photographic clip: the sequence can be right, and the palette can still be a compromise.</p>
<p>The page says the tool was built and tested in current Chrome. There is a live “Does your browser support this?” list, checked against the browser you are using right now by asking it directly, not guessed from the browser name. The checks named on the page are GIF files and background encoding (Web Workers).</p>
<p>The usual reasons to do this join are practical. A short story in several scenes can become one sequence that plays through and loops. A greeting can put a message, a picture, and a sign-off in one file. A reaction pack can chain several reactions into one file to send at once. A tutorial can join steps recorded as separate GIFs into one walkthrough. A long banner loop can be extended by adding a second GIF after the first.</p>
<h2>Check the order, the fit, and the loop point</h2>
<p>Before you download, look at three things.</p>
<p>First, the order. The first row is first, and the names in the list are the file names. If the sign-off is in row one, the loop will start on the sign-off. Move it with the arrows rather than re-exporting the sources.</p>
<p>Second, the fit. If a source has a different shape from the output frame, decide whether you are looking at bars or at cropped edges, and whether the background colour is the one you meant to paint behind every frame. Remember that transparency is not kept, and that at the default output size the first file is not the one being scaled.</p>
<p>Third, the timing, including the loop. Each segment should still feel like itself: a slow GIF still slow, a fast one still fast, a hold still held for as long as it was. Then watch the jump from the last GIF back to the first. The tool will play that loop. It will not retime the files to make the seam gentler, and it will not average the speeds.</p>
<p>If those three look right, download the new file. A combined GIF is not a blend of tempos and not a collage of simultaneous clips. It is one loop, one frame size, and a sequence of delays that still belong to the frames that arrived with them.</p>
<p><em>I build LoopGIF.</em></p>
]]></content:encoded></item><item><title><![CDATA[A Reversed GIF Has the Wrong Rhythm If the Holds Stay Behind]]></title><description><![CDATA[You already have the GIF. You only want it to play backwards. The failure is rarely that the frames run in the wrong order. It is that the holds do not…]]></description><link>https://loopgif.hashnode.dev/a-reversed-gif-has-the-wrong-rhythm-if-the-holds-stay-behind</link><guid isPermaLink="true">https://loopgif.hashnode.dev/a-reversed-gif-has-the-wrong-rhythm-if-the-holds-stay-behind</guid><category><![CDATA[gif]]></category><category><![CDATA[animation]]></category><category><![CDATA[Design]]></category><category><![CDATA[Tutorial]]></category><category><![CDATA[webdev]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 13:01:40 GMT</pubDate><content:encoded><![CDATA[<p>You already have the GIF. You only want it to play backwards. The failure is rarely that the frames run in the wrong order. It is that the holds do not move with them. A pause that belonged to a finished line of type now sits on an empty frame. The motion is reversed. The rhythm is not.</p>
<p><em>I build LoopGIF.</em></p>
<p>There is a second surprise, and it shows up only when the file repeats. A reversed GIF plays backwards once and then jumps back to its new first frame, which is the old last frame. If the original was a one-way motion, such as something falling, that cut looks abrupt. People often expect a bounce: the fall, then the rise, then the fall again, with no snap. That is a different kind of loop. Reverse is not a forward-and-back trip.</p>
<h2>The hold has to travel with the frame</h2>
<p>Each frame of a GIF carries its own delay. A hold is not a global pause overlaid on the file. It is one frame, shown for longer than its neighbours. When you reverse the order, that frame has to take its delay with it. A frame that was on screen for a second is still on screen for a second, only now it appears at another moment. That is what keeps a reversed typing GIF, or any GIF with a hold, looking like itself played backwards.</p>
<p>The live page publishes a typing-GIF example that makes this visible. I did not re-export it; these are the page’s figures. Played backwards, the typing GIF erases itself. The long pause that used to sit on the finished text now comes at the start, because that frame is now first and keeps its delay. The file is listed as 14 frames, 14 KB becoming 15 KB, at 480 × 270.</p>
<p>There is another way to reverse a GIF, and it is the one that breaks the rhythm: reverse the frames, but leave the delays in their old places. The pictures run backwards. The timing map does not. A hold that belonged to a finished word can land on an empty canvas, or a quick in-between frame can inherit a long pause it never earned. <a href="https://loopgif.net/reverse-gif">LoopGIF’s reverse GIF tool</a> does not do that. The frames run from last to first, and each frame keeps its own delay, so the last frame comes first and holds for as long as it did before.</p>
<p>Leave the speed at 1 times if you want that timing unchanged, only reversed in order. That is the default. The preview is built to show this before you write a new file: it plays your GIF from the last frame to the first, with the original delays. Watch where the holds land. If a pause has moved to a frame that should not be held, you are not looking at a faithful reverse.</p>
<h2>Reverse plays backwards once, then jumps</h2>
<p>A reversed GIF plays backwards and then jumps back to its first frame. That first frame is the old last frame. If the original was a one-way motion, such as something falling, the repeat looks abrupt: the object finishes rising, and then it is suddenly back at the end of the fall, ready to rise again. The jump is not a defect in the reverse step. It is what “play backwards, then start over” means.</p>
<p>This is easy to confuse with a boomerang loop. A boomerang plays forward and then backward, so a fall would rise again along the same path and meet itself without a snap. Reverse is not that. It does not play forward and back. It plays the reversed sequence once, then repeats from the start of that reversed sequence.</p>
<p>If you want a rewind joke, or a one-shot undoing of a motion, the jump at the repeat may be acceptable, or you may only play the file once. If you want a loop that can sit on a page forever, look at the original first. If that loop already joins up smoothly, the reverse usually does too. A spinning wheel reversed is a wheel spinning the other way, and the seam is as clean as it was. A one-way path will still be a one-way path, only travelled in the other direction, and the cut back to the start will still show.</p>
<p>The page’s bounce example is the “undoing a motion” case. The bounce plays from its last frame to its first, so the word rises where it fell. The 16 frames and their 5 cs delays — delays measured in hundredths of a second — are the same, only in the opposite order. The file is listed as 34 KB at 480 × 270, unchanged after the reverse. Watch that kind of file at the loop point, not only through the first backward pass. The pass can look perfect and the repeat can still jump.</p>
<h2>What the page’s other examples show</h2>
<p>The same rule — last frame to first, each delay travelling with its frame — shows up in the other published examples. Again, these are the page’s figures.</p>
<p>A banner that scrolled to the left now scrolls to the right. The 39 frames keep their 4 cs delays. The file is listed as 167 KB becoming 168 KB, at 468 × 60. Direction changes; the step timing does not.</p>
<p>There is also a speed example, which is a secondary control, not the reason to use the tool. Reverse at 2 times: every delay is halved, from 10 cs to 5 cs, so the whole GIF lasts 0.8 seconds instead of 1.6. That file is listed as 16 frames, 584 KB, at 320 × 320, with the size unchanged. The order is still reversed. The delays still belong to their frames. They have only been scaled.</p>
<p>Those cases cover the usual reasons to reverse a file you already have: a rewind, an undone motion, text that erases itself, a banner that should travel the other way, or an export written in the wrong order. Judge the preview on two things: whether the holds still land on the frames that earned them, and whether the jump at the repeat is acceptable.</p>
<h2>Speed is optional; size, frames, and transparency stay</h2>
<p>You can change the speed if you like, from a quarter to four times. Leave it at 1 times to keep the timing as it was. Delays are whole hundredths of a second and never below 2, so a fast GIF cannot be made much faster. The order will still reverse, but the delays may not shrink as much as the multiplier suggests.</p>
<p>The result keeps the size, the frame count, and the transparency of the original. Transparent GIFs stay transparent, and soft edges are blended with white. The tool writes a new GIF with a shared palette of up to 255 colours, so the file size can differ a little from the original and there can be small colour changes. The frames are not blurred or resized. The original file is not changed. The After pane shows both sizes.</p>
<p>The tool accepts a GIF only, up to 25 MB and up to 300 frames. Nothing is uploaded. The result is capped at 300 frames, 30 seconds, and 1024 px on the longer side. It is free, needs no account, and runs in the browser: the file is opened on your device and never leaves it. Check the preview, change the speed only if you want to — and only after you have looked at the rhythm at 1 times — then press Reverse GIF. The new file appears next to the preview with its frame count, size, and dimensions, and Download saves it.</p>
<p>A note on browsers, because this page is narrower than some others: the tool was built and tested in current desktop Chrome. It uses ordinary canvas, workers, and GIF decoding. The page has not tested other browsers itself. There is a live “Does your browser support this?” checklist, checked against the browser you are using by asking it directly, not guessed from the browser name.</p>
<h2>Watch the hold, then watch the join</h2>
<p>Drop the file. Look at the preview. Does the hold still sit on the frame that should hold? If the original paused on finished text, the reverse should pause on that same finished text, now at the start. If the rhythm feels off, set the speed back to 1 times and check again. This tool is built not to leave the delays behind.</p>
<p>Then watch the repeat. If the original was already a closed loop — a wheel, a cycle, a banner whose first and last frames already agreed — the reverse should join as cleanly as the source did, only travelling the other way. If the original was a one-way motion, expect a jump back to the old last frame. That jump is the difference between reverse and a forward-and-back loop. The reverse step will not soften it for you.</p>
<p>If both answers look right, download the new file and leave the original alone. The original is not overwritten. A reversed GIF is the same animation, in the opposite order, with the same holds attached to the same frames. The rhythm stays true when the delays travel. The loop stays honest when you do not pretend a one-way path will bounce.</p>
<p><em>I build LoopGIF.</em></p>
]]></content:encoded></item><item><title><![CDATA[A Scrolling Banner That Jumps at the Wrap Is Worse Than a Still]]></title><description><![CDATA[A still banner is honest. The type sits there, and nobody is waiting for it to meet itself. A scrolling banner that almost loops is a different object: it…]]></description><link>https://loopgif.hashnode.dev/a-scrolling-banner-that-jumps-at-the-wrap-is-worse-than-a-still</link><guid isPermaLink="true">https://loopgif.hashnode.dev/a-scrolling-banner-that-jumps-at-the-wrap-is-worse-than-a-still</guid><category><![CDATA[gif]]></category><category><![CDATA[animation]]></category><category><![CDATA[typography]]></category><category><![CDATA[Design]]></category><category><![CDATA[Tutorial]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 12:28:31 GMT</pubDate><content:encoded><![CDATA[<p>A still banner is honest. The type sits there, and nobody is waiting for it to meet itself. A scrolling banner that almost loops is a different object: it promises continuous motion, then hitches when the message wraps. That hitch is the moment the file admits it was assembled, not designed, and it happens on every cycle.</p>
<p><em>I build LoopGIF.</em></p>
<p>This fails on shop-hour strips, stream headers, and newsletter bars. The letters are sharp. The colour is right. Then the last word jumps a few pixels, or a gap collapses, or the first letter slams in a beat early, and the strip looks cheap. A still would have been quieter. The almost-loop is louder than the message.</p>
<h2>The seam is what people notice</h2>
<p>A marquee is a periodic motion. Eyes lock onto the repeat. If the end of the line does not meet the beginning at the same pace, with the same gap, the error is visible even when it is only a handful of pixels. You only need to watch the file loop twice.</p>
<p>That is why “close enough” is a bad standard here. A timeline trimmed by feel, or an animation whose duration does not quite match the travel distance, will often look fine on the first pass and wrong on the second. The first pass has no wrap to judge. The loop does.</p>
<p>The fix is not a fade, a longer pause, or a faster speed meant to hide the join. Those are cosmetics. The join has to be exact: when the message wraps, the frame you land on has to be the same as the first frame, and the distance travelled in between has to be an exact multiple of the step the letters take each frame.</p>
<h2>The travel distance has to be a multiple of the step</h2>
<p>A scrolling text GIF is a sequence of stills. Each frame moves the type a fixed number of pixels. If the total distance — the length of the message plus the gap between repeats — is not a multiple of that step, the cycle cannot close. You either overshoot, undershoot, or invent a last frame that does not belong, and the seam shows.</p>
<p><a href="https://loopgif.net/sliding-text-gif">LoopGIF’s sliding text GIF maker</a> keeps the loop seamless by making that arithmetic true. The distance travelled is an exact multiple of the pixels moved per frame. To make the numbers land, the gap between repeats can grow by a few pixels. You set a spacing; the file may use a slightly larger one. That small adjustment is the difference between a join you can see and a join you cannot.</p>
<p>The other half of the rule is visual. The first frame and the frame after the last are identical, so nothing pops when the GIF starts over. The preview is supposed to loop exactly as the file will, so you can watch the wrap before you download. If you later trim a frame in an editor, or pad the gap by hand without redoing the step size, you can break the multiple and put the jump back in.</p>
<h2>A 150-frame cap, and why a slow long scroll looks steppier</h2>
<p>A GIF is capped at 150 frames. That limit is easy to hit. A long line, set to crawl, wants a large number of small steps. If the tool honoured every requested step, it would need more than 150 frames. Instead, it moves more pixels per frame so the loop still closes inside the cap. The motion is still seamless. It is also a little steppier: you can see the letters jump in slightly larger increments.</p>
<p>That is the trade-off for a very slow, very long scroll. Shorten the message, raise the speed, or accept a slightly chunkier step.</p>
<p>Timing has its own rounding. Delays are whole hundredths of a second and never shorter than two. If the speed you typed does not survive that rounding, the tool tells you the real speed after rounding. Read that number; the preview is already using it.</p>
<h2>The message, the direction, and the canvas</h2>
<p>The message can be up to 200 characters. Line breaks and emoji work; emoji are drawn with your device’s emoji font, so they can look slightly different on someone else’s phone or laptop. For left and right scrolling, the line runs as a single row. For up and down, the text wraps to the canvas width, which is what you want for a short credits roll or a two-line closing message.</p>
<p>Direction is Left, Right, Up, or Down. Speed is in pixels per second. The space between repeats is a separate control.</p>
<p>Canvas presets on the page are 728 × 90, 468 × 60, 640 × 160 (4:1), 600 × 200 (3:1), and 480 × 480 (1:1). Custom sizes go up to 728 pixels wide, 640 tall, and 307,200 pixels in all. Those are the tool’s limits, not a promise that a particular site or app will accept the file. Platforms set their own rules, and those rules change. Check the size your platform asks for, then pick a matching preset or a custom size inside the cap.</p>
<p>Style presets — LED sign, Scroll sign, Ad banner, News ticker — are starting points for colour, type, and layout. You can still change the font, size, alignment, padding, colours, text shadow, and whether the file loops forever. There are thirteen bundled fonts, all under the SIL Open Font License, loaded on demand from the site: Anton, Oswald, Bebas Neue, Pacifico, Dancing Script, JetBrains Mono, VT323, Great Vibes, Lobster, Caveat, Fredoka, Press Start 2P, and Orbitron. Pixel faces such as VT323 and Press Start 2P read as a dot-matrix strip; a condensed display face such as Bebas Neue suits a short ad bar.</p>
<p>The page publishes four worked examples, with settings and export figures. I did not re-export them; these are the page’s own numbers. An LED sign preset in VT323 at 64 px, amber on black, scrolling left at 220 pixels per second on a 480 × 120 canvas, is listed as 57 frames, 215 KB, and a 2.3 s loop. An ad banner at 468 × 60, Bebas Neue on lime, is 39 frames, 167 KB, and 1.6 s. A scroll sign in white Oswald on red, with a text shadow, is 30 frames, 114 KB, and 1.2 s. A direction-up example, a two-line message rolling like end credits on a 320 × 240 canvas, is 38 frames, 215 KB, and 1.5 s. They are useful as a sense of scale, not as targets you have to hit.</p>
<h2>Every frame changes, so the file grows</h2>
<p>A still banner is one image. A scrolling banner redraws the type in a new position on every frame, so almost nothing can be reused. File size grows with the canvas area and the number of frames. Wide leaderboard strips and long, slow loops are the expensive combinations.</p>
<p>The practical cuts are the ones the page itself recommends:</p>
<ul>
<li>a smaller canvas</li>
<li>a shorter message</li>
<li>a faster speed</li>
</ul>
<p>Faster speed usually means fewer frames to cover the same distance, which is why it helps. The estimate under the preview updates as you change settings, but it is an estimate. The file is measured after you download it. Trust the downloaded size when you are checking a platform limit.</p>
<p>Everything runs in the browser. Your text is drawn on a canvas and encoded in a worker on your device when you click Download. Nothing is uploaded, no account is needed, and no watermark is added. It works in current Chrome, Edge, Firefox, and Safari.</p>
<p>If you need the strip to sit on a coloured page without a box, tick Transparent background and set the edge colour to the colour of that page. GIF transparency is on or off — there is no partial alpha — so soft edges of the letters blend toward the edge colour you choose. A transparent banner that jumps at the wrap is still a jumping banner.</p>
<p>There is an optional AI-styled version of the same words. It needs a signed-in account and costs one credit per AI GIF. The free scrolling tool stays free.</p>
<h2>Watch the wrap before you post the file</h2>
<p>Type the message. Pick a direction and a speed. Choose a canvas that matches what the destination actually asks for. Set a gap between repeats that looks like a breath, not a hole. Then leave the preview running and look only at the moment the end meets the start. If that join is clean, and the first frame would be a fair still if someone paused the file, you are ready to download.</p>
<p>If the motion looks steppy, you have probably asked for a long line at a low speed and hit the 150-frame cap. Shorten, speed up, or accept the larger step. If the file is heavier than you can post, shrink the canvas, cut the copy, or raise the speed, then download again and read the measured size — not the estimate.</p>
<p>A still banner does not have this problem, because it never has to meet itself. A scrolling banner does. The work is to make that meeting exact, not to hide it.</p>
<p><em>I build LoopGIF.</em></p>
]]></content:encoded></item><item><title><![CDATA[A Meme Caption That Looks Fine on One Frame Can Fail on the Loop]]></title><description><![CDATA[Screenshotting the punchline flattens the motion. Overlay the text, then check safe margins, contrast on busy frames, and whether the line stays readable…]]></description><link>https://loopgif.hashnode.dev/a-meme-caption-that-looks-fine-on-one-frame-can-fail-on-the-loop</link><guid isPermaLink="true">https://loopgif.hashnode.dev/a-meme-caption-that-looks-fine-on-one-frame-can-fail-on-the-loop</guid><category><![CDATA[gif]]></category><category><![CDATA[meme]]></category><category><![CDATA[Design]]></category><category><![CDATA[Tutorial]]></category><category><![CDATA[Web Development]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 11:58:22 GMT</pubDate><content:encoded><![CDATA[<p>A reaction GIF is a loop first and a joke second. A slow nod, a blink, a wince — the motion is what you are reacting with. The usual way this gets ruined is almost reasonable: you pause on the funniest frame, take a screenshot, and drop that still into a static meme generator. The caption is sharp. The shrug is gone.</p>
<p><em>I build LoopGIF.</em></p>
<p>The other failure is subtler. You do keep the animation, but you treat the caption as a sticker judged on frame one. Then the loop continues: a highlight washes under the letters, a face moves into the top third, a rounded chat bubble clips the last word, or a two-line setup never has time to be read before the file starts again. The first frame looked fine. The GIF, as posted, does not.</p>
<h2>A still of the punchline is a different object</h2>
<p>A still meme quotes one moment. A captioned reaction GIF has to keep happening. If the humour is the pause before a nod, or eyes that close and open, flattening the file to a single frame throws away the timing you were trying to caption.</p>
<p>So the job is not “make a meme.” It is: leave the frames and their delays alone, and add type as a layer on top of them. The motion stays a motion. The text has to work as a companion to that motion, not as a replacement for it.</p>
<p>That also means you should not “fix” a clip by re-encoding it into a new, even timeline just to print a sentence on it. If each frame already has its own delay, keep that. A slow reaction that suddenly plays at a uniform rate is a different clip.</p>
<h2>Leave a margin, even if the classic layout hugs the edge</h2>
<p>Classic GIF memes put one short line near the top and one near the bottom, in a heavy condensed face, white letters with a thick black outline. That layout is not only a style. It leaves the middle of the frame — where the face or the gesture usually lives — comparatively clear.</p>
<p>It is still easy to crowd the edges. A file is rarely shown edge-to-edge. Chat apps round the corners. Some clients overlay a timestamp, a reaction bar, or a compressed preview. A slightly tighter crop on mobile can take the last few pixels off each side. Text that is flush with the bitmap border looks designed in the editor and clipped in the thread.</p>
<p>A practical habit: place the lines near the top and bottom, then nudge them inward until you can still see a strip of picture around the letters. If you are working in percentages of the frame, small nudges are enough; you do not need a new layout. Long captions are the other margin problem. A line that wraps into two or three rows starts eating the subject you were trying to keep visible. Two to five words is usually plenty for a reaction. If the joke needs a paragraph, the paragraph probably belongs next to the GIF, not inside it.</p>
<h2>Contrast has to hold on the busy frames, not the quiet ones</h2>
<p>White fill and a black outline exist because a looping background will not stay one colour. A line that is readable on a dark sweater can disappear when a lamp, a white wall, or a flash of UI passes underneath. The outline is doing most of the work, but it is not magic: a busy, high-frequency patch — confetti, foliage, already-on-screen typography — can still break up the stroke.</p>
<p>Judge contrast by playing the loop, not by pausing on a convenient frame. Watch for the worst frame, the one where the letters sit on the lightest or the most detailed patch. If the subject itself moves into the caption band, move the layer, not the joke. Changing colour is allowed, but swapping the classic white-on-black for a faint tint usually makes the next busy frame worse, not better.</p>
<p>If the source already has text in it — a screenshot of a chat, a slide, a sign — do not park a new caption in the same band. You will get two competing typographic layers, and neither will be easy to read at a glance.</p>
<h2>The line has to stay readable for the length of the loop</h2>
<p>Readability is also a timing problem. A short reaction loop does not give you a poster’s worth of attention. If the file repeats every couple of seconds, a long sentence is still mid-parse when the first frame comes back. Viewers do not pause a GIF in a chat the way they study a still.</p>
<p>Two useful checks:</p>
<ul>
<li>Can you read the caption once, comfortably, in a single pass of the loop, at the size you will actually post? A file exported at 320 pixels wide is the test, not only the large preview.</li>
<li>Does every word need to be on screen for the whole cycle? A setup and a punchline often work better as two layers with different first and last frames — the setup on the first half, the payoff on the second — than as a permanent block of text covering the motion.</li>
</ul>
<p>The opposite mistake is a caption that appears for a handful of frames and vanishes. In a tight loop that reads as a flicker, not a reveal. If you time a line to a beat, give it enough frames to be read, then watch the seam: when the file repeats, the setup should not collide with the punchline from the previous pass.</p>
<p>Do this in the preview, all the way around, more than once. A paused first frame will not show you a mid-loop collision or a line that is too brief.</p>
<h2>Caption the file you have, in the browser</h2>
<p>I maintain <a href="https://loopgif.net/meme-gif-maker">LoopGIF’s meme GIF maker</a>. The free editor runs in your browser, needs no account, and does not upload the file. Encoding starts when you click Download. Frame timing and the real size of the GIF are kept, and no watermark is added.</p>
<p>You can pick one of the eight short reaction loops in the gallery — emoji faces, a bear with a heart, a sleepy moon, a pixel campfire, a toy robot, and a dog, all generated for the page and none with text already burned in — or drop in a GIF or MP4 you have the right to use. The gallery does not include well-known meme templates, because those belong to other people. If you want one of those, you bring your own file and you are responsible for the rights to the picture.</p>
<p>The meme preset is the classic look: Anton (an open-licence face bundled with the page), white fill, black outline, uppercase, one line at the top and one at the bottom. Click a line, type the joke, and drag it if it needs to sit elsewhere. With the preview focused, the arrow keys nudge the selected layer by 1 percent of the frame, or 5 percent with Shift. You can add layers, up to eight, of 120 characters each, and change fonts and colours. Font size and outline are percentages of the frame height, so the caption keeps its proportions if you export at 480 or 320 pixels instead of the original width. Long captions wrap to about 94 percent of the frame width, which is a hint to shorten the line rather than a reason to fill that width.</p>
<p>Each caption has a first and last frame, so a setup can occupy the first half of the loop and a punchline the second. Download at the original width, 480, or 320 pixels. GIFs up to 15 MB and 300 frames are accepted. MP4 clips up to 60 MB are accepted, but only the first 15 seconds are used, at no more than 150 frames; the page tells you when a clip was shortened. The output is the original size capped at 800 pixels on the longer side, or the 480 / 320 widths, with the same aspect ratio.</p>
<p>The same page has an AI meme tab if you do not have a loop yet. It needs a signed-in account and costs 1 credit per GIF. You type the caption (1 to 40 characters from letters, digits, spaces, and a short set of punctuation) and a scene, and the model draws a 16-frame GIF at about 512 pixels, with the words rendered in bold white letters with a black outline. Prompts that ask for real people or brand names are not supported. The page is clear that the model can still misplace or crop text, and that a failed job is refunded. Re-exporting an already generated AI meme at another width or speed in the browser is free. The caption editor above that tab stays free. I am not describing a purchase flow; use the AI tab only if you already have an account and a credit.</p>
<h2>A short check before you post</h2>
<ul>
<li>Keep the motion. Do not screenshot the funniest frame and caption the still.</li>
<li>Place the lines near the top and bottom, then pull them in off the raw edge.</li>
<li>Keep the caption short enough to read in one pass of the loop, at the width you will post.</li>
<li>Play the whole cycle. Look for a bright or busy frame that eats the letters, a face that moves into the type, or a line that appears too briefly.</li>
<li>If the joke is a setup and a punchline, time the layers instead of stacking both sentences on every frame.</li>
</ul>
<p>A meme GIF is typography on a moving picture. The first frame is a poor place to sign off on either one.</p>
<p>I build LoopGIF.</p>
]]></content:encoded></item><item><title><![CDATA[The GIF on the Page Is the Wrong Delivery Format — Export a Looping WebM]]></title><description><![CDATA[A muted, looping video can be much smaller than the GIF you designed with, but WebM is a clip, not an image. Check the container, the loop, and what…]]></description><link>https://loopgif.hashnode.dev/the-gif-on-the-page-is-the-wrong-delivery-format-export-a-looping-webm</link><guid isPermaLink="true">https://loopgif.hashnode.dev/the-gif-on-the-page-is-the-wrong-delivery-format-export-a-looping-webm</guid><category><![CDATA[WebM]]></category><category><![CDATA[gif]]></category><category><![CDATA[performance]]></category><category><![CDATA[Web Development]]></category><category><![CDATA[Tutorial]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 11:33:41 GMT</pubDate><content:encoded><![CDATA[<p>The GIF on your landing page or docs was a convenient way to design a silent loop. It is a poor file to serve. If the network panel still shows a <code>.gif</code>, every visitor downloads a sequence of paletted bitmaps, and a preview that felt light in a design tool can weigh several megabytes on the wire. <em>I build LoopGIF.</em></p>
<h2>Why the preview should not stay a GIF</h2>
<p>A GIF is an image format. Each frame is its own indexed picture, with at most 256 colours, and the delays between frames can vary. That is fine while you are iterating on the motion. It is the wrong delivery format on a page you control, for three practical reasons.</p>
<p>First, the compression model does not suit a photo-like UI, a product shot, or any preview with gradients and camera-like detail. Those clips get heavy fast, because there is no modern video encoder sitting behind the <code>.gif</code>. Second, the browser fetches it as an image. You do not get a <code>&lt;video&gt;</code> element’s loop, mute, or inline playback attributes, and you do not get a straightforward way to declare a codec. Third, a GIF loops by itself; that convenience is why it is still pasted into pages, even when a video file would be smaller.</p>
<p>The replacement is not “any video.” For a short, silent loop on the open web, you want a WebM file: a video container whose mandated video codecs, according to MDN’s <a href="https://developer.mozilla.org/en-US/docs/Web/Media/Guides/Formats/Video_codecs">Web video codec guide</a>, are VP9 and VP8. Both are open and royalty-free. VP9 is the one to prefer when the encoder offers it; VP8 is the fallback in the same container. You do not need an audio track for a page preview.</p>
<h2>What the file should actually contain</h2>
<p>Before you convert anything, be clear about the output you will put in the repository or on the CDN:</p>
<ul>
<li>Container: WebM, with a <code>.webm</code> extension and a <code>video/webm</code> MIME type.</li>
<li>Video codec: VP9, or VP8 if VP9 is not available to the encoder.</li>
<li>Audio: none. A silent loop does not need it, and an unmuted track can also block autoplay.</li>
<li>Frame rate: one constant rate, not the GIF’s per-frame delays.</li>
<li>Loop behaviour: the file itself is a clip that ends. The page has to loop it, or the clip has to be written into the file more than once.</li>
</ul>
<p>On the page, a typical markup looks like this:</p>
<pre><code class="language-plaintext">&lt;video autoplay muted loop playsinline&gt;
  &lt;source src="preview.webm" type="video/webm"&gt;
&lt;/video&gt;
</code></pre>
<p>Those attributes are not decoration. MDN’s documentation for the <a href="https://developer.mozilla.org/en-US/docs/Web/HTML/Reference/Elements/video">HTML video element</a> states that <code>loop</code> is a boolean attribute: if it is present, the browser seeks back to the start when the video ends. <code>muted</code> silences the default audio, which matters because modern browsers block autoplay of videos with an unmuted audio track. <code>playsinline</code> asks for playback inside the element’s own area rather than a fullscreen takeover. <code>autoplay</code> starts playback as soon as it can. If you want a GIF-like preview — it starts on its own, it stays quiet, it does not end on the last frame — all four belong on the element.</p>
<p>You can add a <code>codecs</code> parameter to the type string when you know the exact bitstream, for example VP9. For a simple preview, <code>type="video/webm"</code> is enough to tell the browser this is not an MP4.</p>
<h2>The loop has to be arranged twice</h2>
<p>This is the part people miss when they treat WebM as a drop-in GIF.</p>
<p>A GIF image repeats for as long as it is on the screen. A WebM is just a clip. If you convert a three-second GIF and drop the result into a player with no <code>loop</code> attribute, it plays once and stops. On a page you write, the <code>loop</code> attribute is the real fix. That is the check to make in the markup, not in the file name.</p>
<p>There is a second, complementary check in the file itself. If someone opens the WebM in a desktop player, or you paste it into a context that will not honour <code>loop</code>, a three-second clip still ends. Repeating the animation inside the file — writing the same pass two, three, or more times — makes the clip last longer without any help from the page. That does not replace the <code>loop</code> attribute on your site. It is a safety net for every other place the file might be opened.</p>
<p>Very short source GIFs need a third look at duration. A clip under a second can feel like a blink in a video element. A converter that pads a sub-second pass until the video is at least a second long is doing you a favour, as long as you read the planned length before you click convert.</p>
<h2>Transparency does not travel with this kind of WebM</h2>
<p>A logo or sticker GIF often has a transparent background. That does not automatically become a transparent video.</p>
<p>Two separate facts matter here. The first is about the file you are likely to export: a typical in-browser GIF-to-WebM conversion writes video without an alpha channel and fills transparent pixels with a solid colour you choose. If you leave the fill at white and the page background is cream or dark, you get a box. Pick the colour of the surface where the video will sit.</p>
<p>The second is about playback, even if you later obtain a WebM that does include alpha. MDN’s codec guide notes that Safari does not support alpha transparency in VP9 or in VP8. A “clear” logo video that works in one browser can still show an opaque frame in another. For a landing page or docs preview, a matched solid background is the reliable approach. If the design truly cannot live without see-through pixels, leave that asset as a GIF — or as a different format that your target browsers actually composite — and convert the opaque, photo-like previews instead.</p>
<h2>Convert the GIF and read the plan before you download</h2>
<p>I maintain <a href="https://loopgif.net/gif-to-webm">LoopGIF’s GIF to WebM converter</a>. It is a free in-browser tool: no sign-up, and the GIF is not uploaded. The browser encodes the picture as VP9, or as VP8 when VP9 is not available, and you see the GIF’s size next to the video’s. The page accepts a GIF of up to 25 MB, 300 frames, and 30 seconds of playing time.</p>
<p>The steps are short, but each setting is a check, not a default to ignore.</p>
<ol>
<li>Drop the <code>.gif</code> onto the page, or choose it from disk. The left pane plays the original.</li>
<li>Read the plan under the settings. One line gives the video size, the frame rate, the number of frames, and the length the WebM will have. That length is the one to believe, not the GIF’s own duration.</li>
<li>Confirm the frame rate. A video has one rate; a GIF does not. The tool takes the typical delay in your GIF and picks the nearest of 10, 12, 15, 20, 25, and 30 frames per second, then holds each GIF frame for as many video frames as that delay needs. A 100 ms frame at 20 fps becomes two video frames. Uneven delays are smoothed to that steady rate, which is what a video editor and a <code>&lt;video&gt;</code> element expect.</li>
<li>Set how many times the clip is written into the file, from one to ten. Use this together with <code>loop</code> in your markup, not instead of it. If one pass lasts under a second, the tool repeats it until the video reaches one second and tells you so. The total length is capped at 60 seconds.</li>
<li>If the GIF has transparency, set the background colour to the page colour. The default fill is white (<code>#FFFFFF</code>). This page does not write an alpha channel.</li>
<li>Press Convert to WebM. The video appears beside the GIF, with both file sizes, and plays on a loop. Compare the two sizes before you download. A photo-like GIF is often much smaller as VP9; a tiny, flat graphic may already be small enough that video is not a win. Download saves a <code>.webm</code> file.</li>
</ol>
<p>A few encoder limits are worth knowing in advance. Both sides of the picture are made even, and a frame whose long side is larger than 1920 pixels is scaled down. Resize the GIF first if you need an exact output size. There is no audio in the file. Encoding uses your browser’s WebCodecs <code>VideoEncoder</code>. If the browser has no VP9 or VP8 configuration it can use, the page stops and names what is missing. It does not quietly hand you an MP4 or another container. The tool was built and tested in current Chrome; the page also checks, in your own browser, for WebM encoding through WebCodecs and for background encoding in a Web Worker.</p>
<h2>When WebM is the better file — and when the GIF should stay</h2>
<p>Replace the GIF on the page when all of these are true: you control the HTML, the preview is a silent loop, the GIF is heavy (especially if it is photo-like), and you can either live without transparency or fill the background with a colour that matches the layout. A muted, looping <code>&lt;video&gt;</code> element is the usual way to present that clip, and WebM is a container web projects already use for short loops and effects. A steady frame rate also makes the file easier to drop into an editor later.</p>
<p>Leave the GIF in place when the file is already tiny, when the surrounding context will only display an image (some Markdown image embeds and some chat clients will not treat a WebM like an autoplaying GIF), or when you need real transparency and cannot accept a solid fill. And if the browser in front of you cannot encode WebM, do not assume a download will appear anyway — read the error, and try a current Chrome.</p>
<p>The check that settles the size question is always the same: look at the two numbers after convert. “Usually smaller” is not the same as “always smaller.” If the WebM is not lighter, keep the GIF and spend the effort on a tighter crop or a shorter loop instead.</p>
<p>I build LoopGIF.</p>
]]></content:encoded></item><item><title><![CDATA[Turn a Long Demo Recording Into a Short GIF for GitHub Issues and Docs]]></title><description><![CDATA[A 20–60 second MP4 is a fine walkthrough and a bad GIF. Trim to the action, cap width and frame rate, and leave the long take as video.]]></description><link>https://loopgif.hashnode.dev/turn-a-long-demo-recording-into-a-short-gif-for-github-issues-and-docs</link><guid isPermaLink="true">https://loopgif.hashnode.dev/turn-a-long-demo-recording-into-a-short-gif-for-github-issues-and-docs</guid><category><![CDATA[GitHub]]></category><category><![CDATA[gif]]></category><category><![CDATA[Tutorial]]></category><category><![CDATA[documentation]]></category><category><![CDATA[video]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 10:58:32 GMT</pubDate><content:encoded><![CDATA[<p>You already have the recording: a 20 to 60 second MP4 of a flow you walked through once. Now you want a looping preview in a GitHub issue, pull request comment, or docs page. Converting the whole file is the wrong first step. A GIF is a poor container for a long walkthrough. The preview that helps is usually the two to five seconds in which the thing happens.</p>
<p>I build LoopGIF. The file-size figures below come from a 36-second test clip I encoded myself with FFmpeg 7.1.5 on 17 November 2026 (Asia/Shanghai). GitHub’s attachment and repository upload limits are from documentation I fetched the same day.</p>
<h2>The limit depends on where the file goes</h2>
<p>If you drag a file into an issue, a pull request, or a comment, GitHub’s <a href="https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/attaching-files">Attaching files</a> documentation lists these maximums:</p>
<ul>
<li>10MB for images and gifs</li>
<li>10MB for videos uploaded to a repository owned by a user or organization on a free GitHub plan</li>
<li>100MB for videos uploaded to a repository owned by a user or organization on a paid GitHub plan</li>
<li>25MB for all other files</li>
</ul>
<p>Supported media types in all contexts include PNG, GIF, JPEG, SVG, and video as <code>.mp4</code>, <code>.mov</code>, and <code>.webm</code>. GitHub notes that video codec support is browser-specific and recommends H.264 “for greatest compatibility.”</p>
<p>A GIF in Markdown image syntax loops by itself, with no click. A video attachment plays in an inline player. If you need the whole walkthrough, or narration, keep the MP4: GIF cannot carry audio.</p>
<p>If you are committing the file into the repository for a README or a docs page, a different limit applies. GitHub’s <a href="https://docs.github.com/en/repositories/working-with-files/managing-files/adding-a-file-to-a-repository">Adding a file to a repository</a> documentation says files added via the browser are limited to 25 MiB each, and you can add larger files, up to 100 MiB each, via the command line. That is not a 10MB GIF cap, but a 40-megabyte autoplaying preview is still a bad commit: everyone who opens the page pays for the download.</p>
<h2>I converted a 36-second walkthrough so the sizes are real</h2>
<p>I encoded a 36-second, 1280×720, 30 fps H.264 clip that stands in for a too-long settings-panel demo: a wallpaper, a small animated preview card, a toggle that flips at 8.5 seconds, a scrolling list I left running, and light per-frame grain. The MP4 was 2,836,035 bytes (2.70 MiB). Sizes below are in mebibytes (1024² bytes) and come from this clip only.</p>
<p>A naive conversion of the whole file:</p>
<pre><code class="language-plaintext">ffmpeg -i walkthrough-36s.mp4 -an naive-full.gif
</code></pre>
<p>produced 46,303,267 bytes (44.16 MiB) — 1280×720, 1,080 frames, 36 seconds. That will not attach to a GitHub comment.</p>
<p>I then used FFmpeg’s <code>palettegen</code> and <code>paletteuse</code> filters with Bayer dither, at 10 fps and 640 pixels wide. The full 36 seconds was still 25,797,702 bytes (24.60 MiB): 640×360, 360 frames. The first 30 seconds of that recipe — 300 frames, as long as a 10 fps GIF can be under a 300-frame cap — was 21,088,086 bytes (20.11 MiB). It would pass a 25 MiB repository upload in the browser, and would still be rejected as a 10MB issue attachment.</p>
<p>The four seconds in which the toggle actually happens, from 8.00 s to 12.00 s, are a different story:</p>
<ul>
<li>10 fps, 480 pixels wide: 1,354,185 bytes (1.29 MiB), 480×270, 40 frames</li>
<li>10 fps, 640 pixels wide: 2,274,050 bytes (2.17 MiB), 640×360, 40 frames</li>
<li>12 fps, 480 pixels wide: 1,537,666 bytes (1.47 MiB), 480×270, 48 frames</li>
</ul>
<p>All three sit under the 10MB attachment cap. Trim did the work that shrinking the width and frame rate of the full recording could not. I ran these conversions with FFmpeg, not with the browser tool below.</p>
<p>The GIF ballooned because of the format: an MP4 stores a few full pictures and then the changes between them, while a GIF stores a separate indexed picture of up to 256 colours for each frame. Wallpaper, a moving preview, grain, and a scrolling list leave almost no unchanged pixels to skip.</p>
<h2>Trim first, then cap duration, width, and frame rate</h2>
<p>Watch the recording once and mark the one action the preview has to prove. If that takes more than a sentence, make two short clips, or leave it as video.</p>
<p>I maintain a free <a href="https://loopgif.net/video-to-gif">video to GIF converter</a>. It runs in your browser, needs no sign-up, and does not upload the file: the video is decoded on your device. It accepts MP4, MOV, or WebM up to 300 MB.</p>
<ol>
<li>Drop the recording on the page.</li>
<li>Move the Start and End sliders under the thumbnail strip. The bright part of the strip is what becomes the GIF, and the summary shows the frame count and playing time. Aim for the two to five seconds that carry the point.</li>
<li>Pick a frame rate of 10, 12, or 15 frames per second, a width of up to 640 pixels, a speed from 0.5× to 2×, and Balanced or Smaller file. The page’s own guidance for most clips is 10 to 12 fps at 320 to 480 pixels; use 15 for fast motion and 640 only when the picture needs the detail. Width does the most for file size: a 320 px GIF has less than half the pixels of a 480 px one. Smaller file uses fewer colours. Speed changes how much of the video each frame covers, not the frame rate.</li>
<li>Press Convert to GIF. The result appears with its size. Download saves it.</li>
</ol>
<p>Exports are capped at 300 frames and 30 seconds of playing time. At 10 fps that is a 30-second clip; at 15 fps it is 20 seconds. A 60-second recording cannot be exported whole. When a trim is over the cap, the page shows the numbers and a button that shortens the clip to fit, rather than cutting it silently.</p>
<p>GIF cannot carry audio. Each frame has at most 256 colours and no partial transparency, so gradients can band. Height follows the video’s own shape, including iPhone rotation. The tool was built and tested in current Chrome; decoding uses WebCodecs, with one fallback attempt through a video element. Look at the size before you leave the page: as the 30-second version of my test shows, hitting the frame cap is not the same as fitting under GitHub’s 10MB attachment limit.</p>
<h2>When the MP4 should stay an MP4</h2>
<p>Leave the recording as video when you need the whole 20 to 60 seconds, when you need narration or any other audio, or when the footage is a camera feed, a video call, or anything photo-like that will band and bloat as a GIF.</p>
<p>Attach the MP4, MOV, or WebM to the issue or pull request instead. On a free-plan repository the video cap is 10MB; on a paid plan it is 100MB, per the same attaching-files page. If a README preview is still a multi-megabyte GIF after a tight trim, video is often the better long-term file — but that is a different conversion.</p>
<p>For a short, silent, looping preview of a single UI action, a trimmed GIF is the file that plays itself in a comment and stays under the cap.</p>
<p>I build LoopGIF.</p>
]]></content:encoded></item><item><title><![CDATA[Getting a GIF under Discord's 256 KB emoji limit]]></title><description><![CDATA[Discord caps custom emoji at 256 KB and 128×128. Here is how to crop, flatten, and thin frames so an animated emoji actually uploads.]]></description><link>https://loopgif.hashnode.dev/getting-a-gif-under-discord-s-256-kb-emoji-limit</link><guid isPermaLink="true">https://loopgif.hashnode.dev/getting-a-gif-under-discord-s-256-kb-emoji-limit</guid><category><![CDATA[Design]]></category><category><![CDATA[gif]]></category><category><![CDATA[discord]]></category><category><![CDATA[emoji]]></category><category><![CDATA[Tutorial]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 10:18:51 GMT</pubDate><content:encoded><![CDATA[<p>Discord’s own help articles set a hard cap: a custom emoji file must be under 256 KB. For the sharpest result they tell you to upload at up to 128 × 128 pixels, and they also tell you how it will actually be seen — resized to 32 × 32 in chat. Accepted formats in the help centre are JPEG, PNG, GIF and WebP. Names must be at least two characters, and only letters, numbers and underscores.</p>
<p>Animated emoji use those same name and size rules. You upload a GIF the same way as a still picture: Server Settings → Emoji → Upload Emoji, if you own the server or have the Create Expressions permission.</p>
<p>Two Nitro rules are easy to mix up. Nitro or Nitro Basic unlocks the extra 50 animated-emoji slots on a server; the default is 50 static slots. Once an emoji is on that server, any active member can use the custom emoji there. Nitro or Nitro Basic is what lets you send those custom emoji in other servers, group chats and DMs — not a gate on sending them in the server they already live in.</p>
<p>The tight part is the 256 KB budget, not a published frame-count cap. Discord’s developer docs write the same limit as 256 KiB, and they say an upload over that limit is rejected. A 128 × 128 GIF is small, but GIF still stores a palette for every frame. A two-second reaction is usually enough. Grain, camera motion and gradients spend the budget on pixels that vanish at 32 × 32.</p>
<p>What actually gets a clip under the cap:</p>
<ol>
<li>Cut to the gesture, not the lead-in. One to two seconds, looping on the peak.</li>
<li>Crop square and tight. At chat size, a face or a hand reads; a wide shot and small text do not.</li>
<li>Flatten or remove a plain background. When only the subject changes between frames, GIF compression has much less work to do.</li>
<li>If the file is still over 256 KB, reduce colours first, then drop frames. When you thin frames, keep every second or third one and add the dropped delays onto the frame you keep, so the loop lasts as long as it did instead of speeding up.</li>
</ol>
<p>I use a small browser tool for that last mile, the <a href="https://loopgif.net/discord-emoji-maker">LoopGIF Discord emoji maker</a>. It crops a square, resizes to 128 × 128, and thins the animation until the file fits under 256 KiB. The file is opened on your device and is not uploaded.</p>
<p>I build LoopGIF.</p>
<p>Preview the result at something close to 32 × 32, on both a light and a dark background, before you upload it. If the motion turns into a stutter, shorten the clip or crop tighter rather than shipping a chopped loop.</p>
<p>These figures were checked on 17 November 2026 against Discord’s help articles and developer documentation. Limits change; check the current guidelines before you upload.</p>
]]></content:encoded></item><item><title><![CDATA[GIF to CSS sprite sheet: steps() animation with uneven delays]]></title><description><![CDATA[A raw GIF on a page cannot pause on hover or honor prefers-reduced-motion. A PNG sprite sheet plus CSS steps() can — but only if you account for frames…]]></description><link>https://loopgif.hashnode.dev/gif-to-css-sprite-sheet-steps-animation-with-uneven-delays</link><guid isPermaLink="true">https://loopgif.hashnode.dev/gif-to-css-sprite-sheet-steps-animation-with-uneven-delays</guid><category><![CDATA[CSS]]></category><category><![CDATA[webdev]]></category><category><![CDATA[animation]]></category><category><![CDATA[Tutorial]]></category><dc:creator><![CDATA[LoopGIF]]></dc:creator><pubDate>Tue, 17 Nov 2026 08:58:23 GMT</pubDate><content:encoded><![CDATA[<p>A GIF in an <code>&lt;img&gt;</code> tag is easy to drop onto a page, and awkward to control once it is there. Each frame is limited to a 256-colour palette. You cannot pause the animation on hover without JavaScript. And the image keeps looping even when the user has asked for <code>prefers-reduced-motion</code>.</p>
<p>A sprite sheet — every frame drawn into one PNG — plus a CSS <code>steps()</code> animation addresses those three. You show one cell at a time by moving <code>background-position</code>, so you can pause it, stop it, and use a PNG instead of a GIF. The part most write-ups skip is timing: <code>steps()</code> assumes every frame lasts the same amount of time, and many GIFs do not.</p>
<blockquote>
<p>Disclosure: I build LoopGIF.</p>
</blockquote>
<h2>Export a single-row sheet</h2>
<p>You can build the sheet with any tool that gives you a PNG and each frame's delay. I use the <a href="https://loopgif.net/gif-to-sprite-sheet">GIF to sprite sheet converter on LoopGIF</a>. It runs in the browser (nothing is uploaded), accepts a GIF up to 25 MB and 300 frames, and returns a PNG grid, a JSON file with each frame's position and delay, and a CSS snippet.</p>
<p>For the classic <code>steps()</code> CSS in the next section, make a <strong>single row</strong>:</p>
<ul>
<li><strong>Columns:</strong> set this to the frame count. <code>0</code> is also valid — it picks a near-square grid automatically — but a multi-row grid needs a different animation than the one below. A single row is the layout <code>steps()</code> is built for.</li>
<li><strong>Padding:</strong> leave it at <code>0</code>. Padding is useful as gutters when a game engine scales the sheet, because it stops neighbouring frames bleeding into each other. In the simple CSS below it would also change the distance of each step, so the “move by one frame width” math would be wrong. If you do add padding, use the CSS the converter prints rather than the snippet in this post.</li>
<li><strong>Background:</strong> choose <strong>Transparent</strong> if the sprite will sit on more than one page colour. The other option is a solid <strong>Colour</strong> fill for the whole sheet, including any gaps.</li>
</ul>
<p>Create the sheet and download the <strong>PNG</strong> and the <strong>JSON</strong>. Keep the JSON: you will need the delays when the frames are not evenly timed. Copy the CSS from the text box under the result as well, and compare it with what you write by hand.</p>
<p>One limit to watch: the sheet is never more than 4096 px on a side. If <code>frames × frame width</code> would exceed that, every frame is scaled down by the same factor, and the JSON describes the <strong>scaled</strong> sheet, not the original frame size. Use the cell size from the JSON in your CSS.</p>
<h2>Equal delays: one <code>steps()</code> animation</h2>
<p>Worked example: 8 frames, each 96 × 96 px, in one row, so the sheet is 768 × 96 px. Every frame lasts 100 ms (800 ms in total).</p>
<pre><code class="language-css">.sprite {
  --frames: 8;
  --w: 96px;
  width: var(--w);
  height: 96px;
  background: url("wave-sheet.png") 0 0 / calc(var(--frames) * var(--w)) 100% no-repeat;
  animation: wave 800ms steps(8) infinite; /* 8 × 100 ms */
}

@keyframes wave {
  to {
    background-position: calc(-1 * var(--frames) * var(--w)) 0;
  }
}

@media (prefers-reduced-motion: reduce) {
  .sprite {
    animation: none; /* frame 0 stays on screen */
  }
}

.sprite:hover {
  animation-play-state: paused;
}
</code></pre>
<p>The end position is <code>-8 × 96px</code>, not <code>-7 × 96px</code>. <code>steps(8)</code> uses <code>jump-end</code> by default, so it shows the positions at 0/8, 1/8, … 7/8 of the distance and never the final value. Animate to the full sheet width and each frame appears once. If you write <code>steps(7)</code> by mistake, a frame is skipped or the sheet appears to slide. The step count must equal the frame count.</p>
<h2>When frame delays are not equal</h2>
<p>Open the JSON. Each frame entry includes <code>delay</code> in milliseconds, taken from the GIF and rounded to the hundredths of a second the format stores. If every delay is the same, the duration is <code>delay × frame count</code> and the previous snippet is enough.</p>
<p>Often they are not. A loop will hold the first or last pose longer than the in-between frames — for example 80 ms on most frames and 240 ms on the last one. A single <code>steps(n)</code> call cannot represent that.</p>
<p><strong>Option A — split the timeline.</strong> If only the last frame is held, play the regular frames with <code>steps()</code>, then park on the held frame for the rest of the loop. In this example the first seven frames occupy 7 × 80 ms = 560 ms, which is 70% of an 800 ms loop; the last frame is then shown for 240 ms.</p>
<pre><code class="language-css">.sprite {
  width: 96px;
  height: 96px;
  background: url("wave-sheet.png") 0 0 / 768px 96px no-repeat;
  animation: wave-hold 800ms infinite;
}

@keyframes wave-hold {
  /* 0–70%: frames 0–6 (7 × 80 ms) */
  0% {
    background-position: 0 0;
    animation-timing-function: steps(7);
  }
  70% {
    background-position: calc(-7 * 96px) 0;
    animation-timing-function: step-end;
  }
  /* 70–100%: hold frame 7 (240 ms) */
  100% {
    background-position: calc(-7 * 96px) 0;
  }
}
</code></pre>
<p>The split is <code>(sum of delays before the hold) / (total duration)</code>. Keep the <code>prefers-reduced-motion</code> and hover rules from the previous example; they still apply.</p>
<p>If several frames have their own irregular delays, do not try to fold them into one <code>steps()</code> segment. List a keyframe for each frame, with the percentage for frame <em>i</em> equal to the sum of delays before it, divided by the total. Hold each position with <code>steps(1, end)</code> so the image switches rather than slides — that is the form the converter prints when delays differ, or when you used a grid or padding.</p>
<p><strong>Option B — a short script.</strong> If you would rather drive the sheet from the JSON (useful in a game loop, or when every delay is different), move <code>background-position</code> yourself:</p>
<pre><code class="language-js">// delays copied from the JSON (ms per frame)
const delays = [80, 80, 80, 80, 80, 80, 80, 240];
const el = document.querySelector(".sprite-js");
const w = 96;
let i = 0;

function tick() {
  el.style.backgroundPosition = `${-i * w}px 0`;
  const wait = delays[i];
  i = (i + 1) % delays.length;
  if (!matchMedia("(prefers-reduced-motion: reduce)").matches) {
    setTimeout(tick, wait);
  }
}
tick();
</code></pre>
<p>Give <code>.sprite-js</code> the same width, height, and background image as the CSS version. The first call paints frame 0; if the user prefers reduced motion, the loop does not continue.</p>
<h2>Multi-row sheets</h2>
<p>If a single row would go over 4096 px, leave <strong>Columns</strong> at <code>0</code> (or pick a column count that keeps the sheet a convenient shape) and do not use the one-line <code>steps()</code> snippet above. Either copy the CSS the converter prints for that grid, or read each frame's <code>x</code> and <code>y</code> from the JSON and use those values in the script (<code>background-position: ${-x}px ${-y}px</code>) instead of <code>-i * w</code>.</p>
<h2>Checklist</h2>
<ul>
<li><input disabled type="checkbox" /> Step count equals the frame count</li>
<li><input disabled type="checkbox" /> The animation's end position is the <strong>full</strong> sheet width (<code>jump-end</code>)</li>
<li><input disabled type="checkbox" /> Duration equals the sum of the delays in the JSON</li>
<li><input disabled type="checkbox" /> CSS uses the cell size from the JSON if the sheet was scaled</li>
<li><input disabled type="checkbox" /> Padding is 0, or the keyframes include it</li>
<li><input disabled type="checkbox" /> <code>prefers-reduced-motion</code> is handled</li>
</ul>
]]></content:encoded></item></channel></rss>