Skip to content

Latest commit

 

History

History
136 lines (108 loc) · 4.37 KB

explainer.md

File metadata and controls

136 lines (108 loc) · 4.37 KB
Raw

WebRTC-SVC Explainer

Updated: July 2019

Problem and Motivation

Scalable Video Coding (SVC) is increasingly used in realtime communications applications because of the performance and reliability advantages it offers. Today SVC is supported within WebRTC implementations using proprietary mechanisms and as a result, it is not possible to write web applications supporting SVC that interoperate on multiple browsers.

We believe that there is a need for a simple mechanism to extend the WebRTC API to allow an application to discover the SVC capabilities of supported video codecs as well as to configure an encoder to utilize SVC.

Goals

  • To extend WebRTC's RTCRtpCodecCapability dictionary to allow an application to discover the supported SVC modalities (known as scalabiltyModes) for each video codec.

  • To extend WebRTC's RTCRtpEncodingParameters dictionary to enable an application to configure an RTCRtpSender with the scalabilityMode of each video stream to be sent, as well as to retrieve the scalabilityMode of each video stream that is being sent.

Non-goals

  • To enable differential robustness, where retransmission, forward error correction or redundant coding is applied to some scalability layers but not others.

Outstanding issues

  • Whether to enable spatial scalability to utilize encoding parameters so to achieve feature parity with simulcast, as explained in Issue w3c#14

Key use-cases

  • Centralized video conferencing.

Proposed solutions

  1. Addition of a scalabilityModes attribute to the RTCRtpCodecCapability dictionary.
  2. Addition of a scalabilityMode attribute to the RTCRtpEncodingParameters dictionary.

Example of setting a scalabilityMode

// Example of 3 spatial simulcast layers + 3 temporal layers
var encodings = [
  {rid: 'f', scalabilityMode: 'L1T3'},
  {rid: 'h', scaleResolutionDownBy: 2.0, scalabilityMode: 'L1T3'},
  {rid: 'q', scaleResolutionDownBy: 4.0, scalabilityMode: 'L1T3'}
];

// Example of 2 spatial layers (with 2:1 ratio) + 3 temporal layers
var encodings = [
  {scalabilityMode: 'L2T3'}
];

Example of discovering the scalabilityModes on a Scalable Forwarding Unit (SFU)

// RTCRtpReceiver.getCapabilities('video').codecs[] returned by 
// SFU that can only forward VP8 and VP9 temporal scalability modes
 "codecs": [
    {
      "clockRate": 90000,
      "mimeType": "video/VP8",
      "scalabilityModes": ["L1T2","L1T3"]
    },
    {
      "clockRate": 90000,
      "mimeType": "video/VP9",
      "scalabilityModes": ["L1T2","L1T3","L1T2h","L1T3h"]
    }
]

Example of discovering the scalabilityModes on a sender

// Capabilities returned by RTCRtpSender.getCapabilities('video').codecs[]
  "codecs": [
    {
      "clockRate": 90000,
      "mimeType": "video/VP8",
      "scalabilityModes": ["L1T2","L1T3"]
    },
    {
      "clockRate": 90000,
      "mimeType": "video/rtx",
      "sdpFmtpLine": "apt=96"
    },
    {
      "clockRate": 90000,
      "mimeType": "video/VP9",
      "scalabilityModes": ["L1T2","L1T3","L2T1","L2T2","L2T3","L3T1","L3T2","L3T3","L1T2h","L1T3h","L2T1h","L2T2h","L2T3h","L3T1h","L3T2h","L3T3h"]
    },
    {
      "clockRate": 90000,
      "mimeType": "video/rtx",
      "sdpFmtpLine": "apt=98"
    },
    {
      "clockRate": 90000,
      "mimeType": "video/AV1",
      "scalabilityModes": ["L1T2","L1T3","L2T1","L2T2","L2T3","L3T1","L3T2","L3T3","L1T2h","L1T3h","L2T1h","L2T2h","L2T3h","L3T1h","L3T2h","L3T3h","S2T1","S2T2","S2T3","S3T1","S3T2","S3T3","S2T1h","S2T2h","S2T3h","S3T1h","S3T2h","S3T3h"]
    },
    {
      "clockRate": 90000,
      "mimeType": "video/rtx",
      "sdpFmtpLine": "apt=100"
    }
]

WPT tests

A sample WPT test is here: https://chromium.googlesource.com/chromium/src/+/f2e338d626fbb80ac9cd10d0d8200a4323d15c4d/third_party/blink/web_tests/external/wpt/webrtc-svc/RTCRtpParameters-scalability.html

Detailed design discussion

This proposal has evolved from the approach to SVC support provided in the ORTC API.

  • Developers utilizing the ORTC SVC APIs indicated that while it achieved feature parity with simulcast (e.g. ability to active/inactivate layers and handle arbitrary scale ratios), the API was complex and error prone and also could not express KEY and KEY_SHIFT modes.

As a result of this feedback, a new proposal was developed, focussing on pre-canned scalability structures known as scalability modes.