Susegad UI Accessibility GitHub

How this site looks

The view settings need JavaScript. Every page reads well without them.

Mood

The same pages, in three moods.

Theme

Foundations

Export

A scene (or any canvas) to PNG, WebM and animated GIF.

Open the live demo npx @susegad/cli add export

A scene (or any canvas) to PNG, WebM and animated GIF, for social posts and for the sample video packages/player's demo and checks use — the library dogfoods itself; nothing here fetches anything from the web.

import { canvasToPng, recordSceneToWebm, recordSceneToGif } from './index.js';
import kolam from '../scenes/kolam/index.js'; // the scene's own default export, from defineScene()

const png = await canvasToPng(myCanvas);
const webm = await recordSceneToWebm(kolam, { register: 'warm', seed: 7, durationSec: 3, fps: 24 });
const gif = await recordSceneToGif(kolam, { register: 'warm', seed: 7, durationSec: 3, fps: 10 });

GIF, one frame at a time (gif.stream.js)

recordSceneToGif builds the whole clip in one call, which holds the page still for as long as that takes. For a caller that works frame by frame, yields to the page between frames and can stop (the picture postcard does), gif.stream.js is the same format in pieces: createHistogram() takes pixels a frame at a time and cuts one shared palette(256); createIndexer(palette, { dither }) turns each frame's pixels into palette indices, keeping its nearest-colour lookup between frames (indexFrameFast is the one-shot form); createGifEncoder({ width, height, palette }) takes addFrame(indices, delayMs, rect?) and finish(). A frame may cover only a sub-rectangle, so a card whose picture moves under a still name and stamp encodes only the picture after its first frame. It writes the same LZW bytes as gif.core.js (its test decodes them with an independent reader). Pure; runs in Node and the browser.

PNG (png.js)

canvasToPng(canvas) wraps canvas.toBlob. svgToPng(svg, { width, height, scale, background }) rasterises an inline <svg> (serialised, loaded as an Image, drawn to a canvas) — for the SVG scenes and skins that have no canvas of their own. blobToDataUrl(blob) inlines a PNG for a Folio document or a poster.

WebM (webm.js)

recordCanvasToWebm(canvas, { fps, durationSec, draw }) opens canvas.captureStream(0) — a stream that takes no automatic frames — and calls draw(t, i, frames) once per frame, requesting exactly that frame from the track afterwards. The recording is therefore exactly the frames asked for, at exactly the times asked for, never whatever the wall clock managed while MediaRecorder was running: the same seed and frame count give the same video, run to run, however loaded the machine is.

recordSceneToWebm(def, { seed, register, params, fps, durationSec }) drives a scene definition's own model() and createRenderer() directly at fixed, evenly spaced times (never the scene's wall-clock loop), so a scene export is reproducible the same way.

Known limitation: the WebM this writes is not scrubbable in Chromium. MediaRecorder does not write a Cues (seek index) element, so Chromium reports the file's video.seekable as [0, 0] however much is buffered — confirmed directly against a sample clip (packages/player/fixtures, readyState === 4, fully buffered, seekable still [0, 0]). The clip plays straight through correctly and its duration reads correctly once seeked once (the standard workaround), but a person cannot drag a scrubber to an arbitrary point without an external remux (for example ffmpeg -c copy to add Cues). See packages/player/fixtures/README.md for how this shaped that package's own checks. Worth a follow-up if an exported clip ever needs to be scrubbed outside this library.

GIF (gif.js, gif.core.js)

gif.core.js is a small, dependency-free GIF89a encoder: median-cut palette quantisation, nearest-colour indexing, and an LZW core matching GIF's own variable-code-size variant, all pure and tested in Node against a from-scratch reference decoder (not a claim that any particular GIF viewer accepts the output, only that the bytes follow the format's own rules). gif.js is the browser edge: paint a frame, read its pixels with getImageData, repeat, encode. recordSceneToGif mirrors recordSceneToWebm's fixed-time driving of a scene.

Budget

18 KB of behaviour source for the whole package; see registry.json for the breakdown. gif.core.js alone is the biggest part (7.9 KB) and is the one piece with real algorithmic weight (median cut, LZW), which is why it is separated out and tested on its own, the way a component's *.core.js is (decision 0010).