webm-wasm

0.4.1 • Public • Published

webm-wasm

webm-wasm lets you create webm videos in JavaScript via WebAssembly. The library consumes raw RGBA32 buffers (4 bytes per pixel) and turns them into a webm video with the given framerate and quality. This makes it compatible out-of-the-box with ImageData from a <canvas>. With realtime mode you can also use webm-wasm for streaming webm videos.

Works in all major browsers (although Safari can’t play webm 🐼).

The wasm module was created by emscripten’ing libvpx, libwebm and libyuv.

$ npm install --save webm-wasm

Note: This is a proof-of-concept and not a production-grade library.

Usage

webm-wasm runs in a worker by default. It works on the web and in in Node, although you need Node 11+ with the --experimental-worker flag.

Quickstart

// 1. Load the `webm-wasm.js` file in a worker
const worker = new Worker("webm-worker.js");
// 2. Send the path to the `.wasm` file
worker.postMessage("./webm-wasm.wasm");
// 3. Wait for the worker to be ready
await nextMessage(worker);
// 4. Send the parameters for the constructor
worker.postMessage({
  width: 512,
  height: 512
  // ... more constructor options below
});
// 5. Start sending frames!
while (hasNextFrame()) {
  // ArrayBuffer containing RGBA24 data
  const buffer = getFrame();
  worker.postMessage(buffer, [buffer]);
}
// 6. Signal end-of-stream
worker.postMessage(null);
// 7. Get the webm file as an ArrayBuffer
const webm = await nextMessage(worker);
// 8. Cleanup
worker.terminate();

(You can find an implementation of nextMessage() in src/worker/webm-worker.js)

Constructor options

  • width (default: 300): Width of the video
  • height (default: 150): Height of the video
  • timebaseNum (default: 1): Numerator of the fraction for the length of a frame
  • timebaseDen (default: 30): Denominator of the fraction for the length of a frame
  • bitrate (default: 200): Bitrate in kbps
  • realtime (default: false): Prioritize encoding speed over compression ratio and quality. With realtime mode turned off the worker will send a single ArrayBuffer containing the entire webm video file once input stream has ended. With realtime mode turned on the worker will send an ArrayBuffer in regular intervals.

From a CDN

Worker code can’t be loaded from another origin directly, even when the source is CORS-enabled. It is, however, still possible to load webm-wasm from a CDN like unpkg.com with a little workaround:

const buffer = await fetch(
  "https://unpkg.com/webm-wasm@<version>/dist/webm-worker.js"
).then(r => r.arrayBuffer());
const worker = new Worker(
  URL.createObjectURL(new Blob([buffer], { type: "text/javascript" }))
);
worker.postMessage("https://unpkg.com/webm-wasm@<version>/dist/webm-wasm.wasm");
// Continue as normal

WebAssembly

If you just want to use the WebAssembly module directly, you can grab webm-wasm.wasm as well as the the Emscripten glue code webm-wasm.js.

The WebAssembly module exposes a C++ class via embind:

class WebmEncoder {
  public:
    // Same options as above. `cb` is a callback function that takes an ArrayBuffer. 
    WebmEncoder(int timebase_num, int timebase_den, unsigned int width, unsigned int height, unsigned int bitrate, bool realtime, val cb);
    bool addRGBAFrame(std::string rgba);
    bool finalize();
    std::string lastError();
    // ... 
} 

Experimental: TransformStreams

Transferable Streams are behind the “Experimental Web Platform Features” flag in Chrome Canary. The alternative webm-transformstreamworker.js makes use of them to expose the webm encoder. Take a look at the demos to see the usage.

Demos

To run the web demos, start the websever using

$ npm run serve

To run the node demos, run them directly (requires Node 11+):

$ node --experimental-worker ./node-simple.js

Building

Because the build process is completely Dockerized, Docker is required for building webm-wasm.

$ npm install
$ npx napa
$ npm run build

Apache 2.0

Readme

Keywords

none

Package Sidebar

Install

npm i webm-wasm

Weekly Downloads

1,034

Version

0.4.1

License

Apache-2.0

Unpacked Size

902 kB

Total Files

29

Last publish

Collaborators

  • surma