/web-demuxer

Demux media files in the browser using WebAssembly, designed for WebCodecs 在浏览器中实现媒体文件的解封装,专为WebCodecs设计

Primary LanguageTypeScriptMIT LicenseMIT

English | 简体中文

Web-Demuxer

Demux media files in the browser using WebAssembly, designed for WebCodecs.

Purpose

WebCodecs only provide the ability to decode, but not to demux. mp4box.js is cool, but it only supports mp4 demux. Web-Demuxer aims to support as many multimedia formats as possible at once.

Features

  • 🪄 Specifically designed for WebCodecs, the API is very friendly for WebCodecs development, you can easily realize the media file demux.
  • 📦 One-time support for a variety of media formats, such as mov/mp4/mkv/webm/flv/m4v/wmv/avi, etc.
  • 🧩 Support for customized packaging, you can adjust the configuration, packaged in a specified format demuxer

Install

NPM

npm install web-demuxer

CDN

<script type="module">
  import { WebDemuxer } from 'https://cdn.jsdelivr.net/npm/web-demuxer/+esm';
</script>

Usage

import { WebDemuxer } from "web-demuxer"

const demuxer = new WebDemuxer({
  // ⚠️ you need to put the dist/wasm-files file in the npm package into a static directory like public
  // making sure that the js and wasm in wasm-files are in the same directory
  wasmLoaderPath: "https://cdn.jsdelivr.net/npm/web-demuxer@latest/dist/wasm-files/ffmpeg.min.js",
})

// Take the example of seeking a video frame at a specified point in time
async function seek(file, time) {
  // 1. load video file
  await demuxer.load(file);

  // 2. demux video file and generate WebCodecs needed VideoDecoderConfig and EncodedVideoChunk
  const videoDecoderConfig = await demuxer.getVideoDecoderConfig();
  const videoChunk = await demuxer.seekEncodedVideoChunk(time);

  // 3. use WebCodecs to decode frame
  const decoder = new VideoDecoder({
    output: (frame) => {
      // draw frame...
      frame.close();
    },
    error: (e) => {
      console.error('video decoder error:', e);
    }
  });

  decoder.configure(videoDecoderConfig);
  decoder.decode(videoChunk);
  decoder.flush();
}

Examples

API

new WebDemuxer(options: WebDemuxerOptions)

Creates a new instance of WebDemuxer.

Parameters:

  • options: Required, configuration options.
    • wasmLoaderPath: Required, the path to the corresponding JavaScript loader file for wasm (corresponding to the ffmpeg.js or ffmpeg-mini.js in the dist/wasm-files directory of the npm package).

    ⚠️ You must ensure that the wasm and JavaScript loader files are placed in the same accessible directory, the JavaScript loader will default to requesting the wasm file in the same directory.

load(source: File | string): Promise<void>

Loads a file and waits for the wasm worker to finish loading. The subsequent methods can only be called after the load method has been successfully executed.

Parameters:

  • source: Required, support the File object or file URL to be processed.
getVideoDecoderConfig(): Promise<VideoDecoderConfig>

Parses the video stream to obtain the VideoDecoderConfig of the file, and the return value can be directly used as an argument for the configure method of VideoDecoder.

getAudioDecoderConfig(): Promise<AudioDecoderConfig>

Parses the audio stream to obtain the AudioDecoderConfig of the file, and the return value can be directly used as an argument for the configure method of AudioDecoder.

seekEncodedVideoChunk(time: number, seekFlag?: AVSeekFlag): Promise<EncodedVideoChunk>

Retrieves the video data at the specified time point (default keyframe), and the return value can be directly used as an argument for the decode method of VideoDecoder.

Parameters:

  • time: Required, in seconds.
  • seekFlag: The seek flag, defaults to 1 (seek backward). See AVSeekFlag for more details.
seekEncodedAudioChunk(time: number, seekFlag?: AVSeekFlag): Promise<EncodedAudioChunk>

Retrieves the audio data at the specified time point, and the return value can be directly used as an argument for the decode method of AudioDecoder.

Parameters:

  • time: Required, in seconds.
  • seekFlag: The seek flag, defaults to 1 (seek backward). See AVSeekFlag for more details.
readAVPacket(start?: number, end?: number, streamType?: AVMediaType, streamIndex?: number, seekFlag?: AVSeekFlag): ReadableStream<WebAVPacket>

Returns a ReadableStream for streaming packet data.

Parameters:

  • start: The start time for reading, in seconds, defaults to 0, reading packets from the beginning.
  • end: The end time for reading, in seconds, defaults to 0, reading until the end of the file.
  • streamType: The type of media stream, defaults to 0, which is the video stream. 1 is audio stream. See AVMediaType for more details.
  • streamIndex: The index of the media stream, defaults to -1, which is to automatically select.
  • seekFlag: The seek flag, defaults to 1 (seek backward). See AVSeekFlag for more details.

Simplified methods based on the semantics of readAVPacket:

  • readVideoPacket(start?: number, end?: number, seekFlag?: AVSeekFlag): ReadableStream<WebAVPacket>
  • readAudioPacket(start?: number, end?: number, seekFlag?: AVSeekFlag): ReadableStream<WebAVPacket>
getAVStream(streamType?: AVMediaType, streamIndex?: number): Promise<WebAVStream>

Gets information about a specified stream in the media file.

Parameters:

  • streamType: The type of media stream, defaults to 0, which is the video stream. 1 is audio stream. See AVMediaType for more details.
  • streamIndex: The index of the media stream, defaults to -1, which is to automatically select.

Simplified methods based on the semantics of getAVStream:

  • getVideoStream(streamIndex?: number): Promise<WebAVStream>
  • getAudioStream(streamIndex?: number): Promise<WebAVStream>
getAVStreams(): Promise<WebAVStream[]>

Get all streams in the media file.

getAVPacket(time: number, streamType?: AVMediaType, streamIndex?: number, seekFlag?: AVSeekFlag): Promise<WebAVPacket>

Gets the data at a specified time point in the media file.

Parameters:

  • time: Required, in seconds.
  • streamType: The type of media stream, defaults to 0, which is the video stream. 1 is audio stream. See AVMediaType for more details.
  • streamIndex: The index of the media stream, defaults to -1, which is to automatically select.
  • seekFlag: The seek flag, defaults to 1 (seek backward). See AVSeekFlag for more details.

Simplified methods based on the semantics of getAVPacket:

  • seekVideoPacket(time: number, seekFlag?: AVSeekFlag): Promise<WebAVPacket>
  • seekAudioPacket(time: number, seekFlag?: AVSeekFlag): Promise<WebAVPacket>
getAVPackets(time: number, seekFlag?: AVSeekFlag): Promise<WebAVPacket[]>

Simultaneously retrieves packet data on all streams at a certain time point and returns in the order of the stream array.

Parameters:

  • time: Required, in seconds.
  • seekFlag: The seek flag, defaults to 1 (seek backward). See AVSeekFlag for more details.
getMediaInfo(): Promise<WebMediaInfo> // 2.0 New

Get the media information of a file, the output is referenced from ffprobe

{
    "format_name": "mov,mp4,m4a,3gp,3g2,mj2",
    "duration": 263.383946,
    "bit_rate": "6515500",
    "start_time": 0,
    "nb_streams": 2,
    "streams": [
        {
            "id": 1,
            "index": 0,
            "codec_type": 0,
            "codec_type_string": "video",
            "codec_name": "h264",
            "codec_string": "avc1.640032",
            "color_primaries": "bt2020",
            "color_range": "tv",
            "color_space": "bt2020nc",
            "color_transfer": "arib-std-b67",
            "profile": "High",
            "pix_fmt": "yuv420p",
            "level": 50,
            "width": 1080,
            "height": 2336,
            "channels": 0,
            "sample_rate": 0,
            "sample_fmt": "u8",
            "bit_rate": "6385079",
            "extradata_size": 36,
            "extradata": Uint8Array,
            "r_frame_rate": "30/1",
            "avg_frame_rate": "30/1",
            "sample_aspect_ratio": "N/A",
            "display_aspect_ratio": "N/A",
            "start_time": 0,
            "duration": 263.33333333333337,
            "rotation": 0,
            "nb_frames": "7900",
            "tags": {
                "creation_time": "2023-12-10T15:50:56.000000Z",
                "language": "und",
                "handler_name": "VideoHandler",
                "vendor_id": "[0][0][0][0]"
            }
        },
        {
            "id": 2,
            "index": 1,
            "codec_type": 1,
            "codec_type_string": "audio",
            "codec_name": "aac",
            "codec_string": "mp4a.40.2",
            "profile": "",
            "pix_fmt": "",
            "level": -99,
            "width": 0,
            "height": 0,
            "channels": 2,
            "sample_rate": 44100,
            "sample_fmt": "",
            "bit_rate": "124878",
            "extradata_size": 2,
            "extradata": Uint8Array,
            "r_frame_rate": "0/0",
            "avg_frame_rate": "0/0",
            "sample_aspect_ratio": "N/A",
            "display_aspect_ratio": "N/A",
            "start_time": 0,
            "duration": 263.3839455782313,
            "rotation": 0,
            "nb_frames": "11343",
            "tags": {
                "creation_time": "2023-12-10T15:50:56.000000Z",
                "language": "und",
                "handler_name": "SoundHandler",
                "vendor_id": "[0][0][0][0]"
            }
        }
    ]
}
setLogLevel(level: AVLogLevel) // 2.0 New

Parameters:

  • level: Required, output log level, see AVLogLevel for details.
destroy(): void

Destroys the instance and releases the worker.

Custom Demuxer

Currently, two versions of the demuxer are provided by default to support different formats:

  • dist/wasm-files/ffmpeg.js: Full version (gzip: 996 kB), larger in size, supports mov, mp4, m4a, 3gp, 3g2, mj2, avi, flv, matroska, webm, m4v, mpeg, asf
  • dist/wasm-files/ffmpeg-mini.js: Minimalist version (gzip: 456 kB), smaller in size, only supports mov, mp4, m4a, 3gp, 3g2, matroska, webm, m4v

If you want to use a smaller size version, you can use version 1.0 of web-demuxer, the lite version is only 115KB
Version 1.0 is written in C, focuses on WebCodecs, and is small in size, while version 2.0 uses C++ Embind, which provides richer media information output, is easier to maintain, and is large in size

You can also implement a demuxer for specific formats through custom configuration:

First, modify the enable-demuxer configuration in the Makefile

DEMUX_ARGS = \
    --enable-demuxer=mov,mp4,m4a,3gp,3g2,mj2,avi,flv,matroska,webm,m4v,mpeg,asf

Then execute npm run dev:docker:arm64 (if on Windows, please execute npm run dev:docker:x86_64) to start the Docker environment.

Finally, execute npm run build:wasm to build the demuxer for the specified formats.

License

This project is primarily licensed under the MIT License, covering most of the codebase.
The lib/ directory includes code derived from FFmpeg, which is licensed under the LGPL License.