This project is provides a minimal implementation (inside the browser) of a live video and audio encoder and video / audio player. The goal is to provide a minimal live platform implementation that helps learning low latency trade offs, and facilitates low latency experimentation.

We also used it to start experimenting in media over QUIC transport (learn more about it in moq IETF workgroup)

Fig1: Main block diagram

For the server/relay side we have used go-media-webtransport-server, a server that we developed media serving purposes using QUIC as underlying transport

This is a more refined packager than v1, it keeps same functionality bit is more optimal in terms of efficiency and also keeping a good trade off with flexibility.

The efficiency of this packager is ~91%, meaning 91% of the bytes sent to transport are payload bytes (1Mbps video and 32Kbps audio)

Fig2.1: Media packager v2

We use a very experimental, flexible, and NON optimal packager that we created for specifically for this, see fig2.2

The efficiency of this packager is ~70%, meaning 70% of the bytes sent to transport are payload bytes (1Mbps video and 32Kbps audio)

Fig2.2: Media packager v1

The encoder is based on Webcodecs, and AudioContext, see the block diagram in fig3

Fig3: Encoder block diagram

Note: We have used WebTransport, so the underlying transport is QUIC, QUIC streams to be more accurate

Video encoding config:

// Video encoder config
const videoEncoderConfig = {
  encoderConfig: {
    codec: 'avc1.42001e', // Baseline = 66, level 30 (see: https://en.wikipedia.org/wiki/Advanced_Video_Coding)
    width: 640,
    height: 480,
    bitrate: 2_000_000, // 2 Mbps
    framerate: 30,
    latencyMode: 'realtime', // Sends 1 chunk per frame
  },
  encoderMaxQueueSize: 2,
  keyframeEvery: 60,
};

Audio encoder config:

// Audio encoder config
const audioEncoderConfig = {
  encoderConfig: {
    codec: 'opus', // AAC NOT implemented at the time I wrote this (it is in their roadmap)
    sampleRate: 48000,
    numberOfChannels: 1,
    bitrate: 32000,
  },
  encoderMaxQueueSize: 10,
};

Muxer config:

const muxerSenderConfig = {
  audioMaxMaxQueueSizeMs: 200,
  videoMaxMaxQueueSizeMs: 100,

  maxInFlightAudioRequests: 100,
  maxInFlightVideoRequests: 50,

  urlHostPort: '',
  urlPath: '',

  maxAgeChunkS: 120,
}

Main encoder webpage and also glues all encoder pieces together

• When it receives an audio OR video raw frame from a_capture or v_capture:

  • Adds it into TimeChecker
  • Sends it to encoder

• When it receives an audio OR video encoded chunk from a_encoder or v_encoder:

  • Gets the wall clock generation time of 1st frame in the chunk
  • Sends the chunk (augmented with wall clock, seqId, and metadata) to the muxer

Stores the frames timestamps and the wall clock generation time from the raw generated frames. That allows us keep track of each frame / chunk creation time (wall clock)

WebWorker that waits for the next RGB or YUV video frame from capture device, augments it adding wallclock, and sends it via post message to video encoder

WebWorker Receives the audio PCM frame (few ms, ~10ms to 25ms of audio samples) from capture device, augments it adding wallclock, and finally send it (doing copy) via post message to audio encoder

WebWorker Encodes RGB or YUV video frames into encoded video chunks

• configures the VideoEncoder in realtime latency mode, so it delivers a chunk per video frame
• Receives the video RGB or YUV frame from v_capture.js
• Adds the video frame to a queue. And it keeps teh queue smaller than encodeQueueSize. It helps when encoder is overwhelmed
• Specifies I frames based on config var keyframeEvery
• It receives a chunk and delivers it to the next stage (muxer)

WebWorker Encodes PCM audio frames (samples) into encoded audio chunks

• Configures the AudioEncoder, it delivers a chunk per audio frame
• Receives the audio PCM frame from a_capture.js
• Adds the audio frame to a queue. And it keeps the queue smaller than encodeQueueSize. It helps when encoder is overwhelmed
• It receives the chunk and delivers it to the next stage (muxer)

WebWorker Sends audio and video chunks to the server / relay

• Opens a WebTransport session against the server
• Creates a Unidirectional (encoder -> server) QUIC stream per every frame
• Receives audio and video chunks from a_encoder.js and v_encoder.js
• Keeps 2 queues of chunks (one per audio other per video)
• For audio: If we try to add a chunk and the queue is over configured size audioMaxMaxQueueSizeMs it drops the 1st element on the queue
• For video: If we try to add a chunk and the queue is over configured size videoMaxMaxQueueSizeMs it drops ALL the queue and continues dropping until we get a key frame. I also waits for videoPenaltyAudioSentFrames audio frames to be sent before adding a new video frame the video queue
• It continuously send audio and video chunk, audio is always sent first if there are any in the audio queue. In other words it only sends video when audio sending queue is empty
• It keeps number of inflight requests always below configured value maxInFlightRequest
• When sending chunks adds the following metadata as inside header fields:

'Cache-Control': max-age=AA`, // Indicates server to cache this data for AA seconds (except init segments)
'Joc-Media-Type': mediaType, // audio, video, or data
'Joc-Timestamp': timestamp, // Chunk PTS in nanoseconds
'Joc-Duration': duration, // Duration in nanoseconds
'Joc-Chunk-Type': chunkType, // key, delta, init
'Joc-Seq-Id': seqId, // Unique and monotonically increasing inside specific media type track
'Joc-First-Frame-Clk': firstFrameClkms, // EPOCH ms when the 1st sample in that chunk was captured
'Joc-Uniq-Id': pId, // Unique request ID (not used)

The player uses Webcodecs and AudioContext / Worklet, SharedArrayBuffer, and Atomic

Note: We have used WebTransport, so the underlying transport is QUIC, quicStreams to be more accurate

Fig4: Player block diagram

To keep the audio and video in-sync, following strategy is applied:

• Audio renderer (audio_circular_buffer.js) keeps track of last played timestamp (delivered to audio device by source_buffer_worklet.js) by using PTS value in the current playing AudioData frame and adding the duration of the number of samples delivered. This information is accessible from player page via timingInfo.renderer.currentAudioTS, who also adds the hardware latency provided by AudioContext.
• Every time we sent new audio samples to audio renderer the video renderer video_render_buffer (who contains YUV/RGB frames + timestamps) gets called and:
  • Returns / paints the oldest closest (or equal) frame to current audio ts (timingInfo.renderer.currentAudioTS)
  • Discards (frees) all frames older current ts (except the returned one)
• It is worth saying that AudioDecoder does NOT track timestamps, it just uses the 1st one sent and at every decoded audio sample adds 1/fs (so sample time). That means if we drop and audio packet those timestamps will be collapsed creating A/V out of sync. To workaround that problem we calculate all the audio GAPs duration timestampOffset (by last playedTS - newTS, ideally = 0 if NO gaps), and we compensate the issued PTS by that.

WebWorker

• Opens webtransport session, sending following params as querystring:
  • old_ms: Time from live edge to start sending (0 = live edge)
  • vj_ms: Video jitter size in ms
  • aj_ms: Audio jitter size in ms
• Waits for incoming unidirectional (Server -> Player) QUIC streams
  • It cancels the stream if NOT finished after quicStreamsExpirationTimeoutMs (Fix to 7000ms for now, use jitter buffer value)
• For every received chunk (QUIC stream) we create and send:
  • Init: Pass the payload as it is
  • Video: Create EncodedVideoChunk
  • Audio: Create EncodedAudioChunk

Since we do not have any guarantee that QUIC streams are delivered in order we need to order them before sending them to the decoder. This is the function of the deJitter. We create one instance per track, in this case one for Audio, one for video video

• Receives the chunks from demuxer_downloader.js
• Adds them into a sorted list, we sort by ascending seqID
• When list length (in ms is > bufferSizeMs) we deliver (remove) the 1st element in the list
• It also keeps track of delivered seqID detecting:
  • Gaps / discontinuities
  • Total QUIC Stream lost (not arrived in time)

WebWorker when it receives and audio chunk it decodes it and it sends the audio PCM samples to the audio renderer. AudioDecoder does NOT track timestamps on decoded data, it just uses the 1st one sent and at every decoded audio sample adds 1/fs (so sample time). That means if we drop and audio packet those timestamps will be collapsed creating A/V out of sync. To workaround that problem we calculate all the audio GAPs duration timestampOffset and we publish that to allow other elements in the pipeline to have accurate idea of live head position

• Receives audio chunk
  • If discontinuity detected (reported by jitter_buffer.js) then calculate lost time by:
    • lostTime = currentChunkTimestamp - lastChunkSentTimestamp; Where lastChunkSentTimestamp = lastSentChunk.timestamp + lastSentChunk.duration
    • timestampOffset += lostTime
• Decode chunk and deliver PCM data + timestampOffset

Leverages SharedArrayBuffer and Atomic to implement following mechanisms to share data in a "multi thread" environment:

• Circular buffer (sharedAudiobuffers): Main buffer used to share audio PCM data from decoder to renderer source_buffer_worklet.js
• State communication (sharedStates): Use to share states and data between renderer source_buffer_worklet.js and main thread

AudioWorkletProcessor, implements an audio source Worklet that sends audio samples to renderer.

• It reads new audio samples from circular buffer
• The samples are played at sampling freq rate
  • In case the buffer is exhausted (underrun) it will insert silence samples and notify play timing according to that.
• Reports last PTS rendered (this is used to keep A/V in sync)

WebWorker, Decodes video chunks and sends the decoded data (YUV or RGB) to the next stage (video_render_buffer.js)

• Initializes video decoder with init segment
• Sends video chunks to video decoder
  • If it detects a discontinuity drops all video frames until next IDR frame
• Sends the decoded frame to video_render_buffer.js

Buffer that stores video decoded frames

• Received video decoded frames
• Allows the retrieval of video decoded frames via timestamps
  • Automatically drops all video frames with smaller timestamps than the requested one

• Every audio received chunk timestamp and clkms (wall clock) is added into audioTimeChecker queue (instance of TimeChecker)
• The renderer.currentAudioTS (current audio sample rendered) is used to get the closest wall clock time from audioTimeChecker
• The UI displays: `Latency = Now - renderedSampleGeneratedWallClock and that is displayed in the UI

Note: Encoder and Player clock have to be in sync for this metric to be accurate. You can use same computer as encoder & player, then metric should be pretty accurate

const downloaderConfig = {
    targetBufferS: 1, // Target player buffer size, we will request EDGE - this value

    urlHostPort: '',
    urlPath: '',
}

• Follow the installation instructions of go-media-webtransport-server
• Clone this repo

git clone [email protected]:jordicenzano/webcodecs-capture-play.git

• Install phyton (see this guide)
• Run local webserver by calling:

./start-http-server-cross-origin-isolated.py 

Note: You need to use this script to run the player because it adds some needed headers (more info here)

• Load encoder webpage, url: http://localhost:8080/src-encoder/
  • Click "start"
• Load player webpage, url: http://localhost:8080/src-player/
  • Copy StreamID from encoder webpage and paste it into Receiver demuxer - streamID
  • Click "start"

ENJOY YOUR POCing!!! :-)

Fig5: Player block diagram

Fig6: Player block diagram

• Encoder: Cancel QUIC stream after some reasonable time (?) in mode live
• Encoder: Implement priorities (sendOrder when available)
• Player: Do not use main thead for anything except reporting
• Player: Accept different sampling frequency than the audio renderer
• Player/server: Cancel QUIC stream if arrives after jitter buffer
• Accelerate playback if we are over latency budget
• All:
  • Accept B frames (DTS)
  • Accept multi track
• TODO:???

webcodecs-capture-play is released under the MIT License.