Position Statements

Topics you are interested in discussing at FOMS:

  • WebVTT?: what features browsers should implement interoperably, now that WebVTT? is in CR
    • Need HTML character reference support in vtt.js (Brion Vibber)
  • Frame Accuracy: frame level APIs? we could implement and the use cases they would enable. Frame-by-frame seek, frame accurate timing information
    Quoting Daiz from Video Dev Slack discussion:
  lack of frame accuracy pretty much kills a lot of potential video applications in the browser
  high quality non-linear editing is another thing where lack of frame accuracy basically kills the whole thing
  pretty hard to try to edit video properly if you can't do the exact accurate cuts you want
  on the subject of browsers and video frame accuracy
  one possibility that came to mind would be that instead of using just the audio for currentTime
  if there's video, currentTime would instead be set to presentation timestamps for currently displayed video frame
  though I can't say how easy or hard that would be to actually implement
  a higher level solution would be to provide new APIs entirely
  like if we had a VideoDecoder API
  which would tap into decoders and provide RGB images with timestamps (could even give raw YUV as an option for anyone wanting to do conversions manually)
  then actual frame display would be up for a JS dev to build
  and since they'd have exact control over frame display, you could naturally do frame accurate things
  it could also allow JS devs to do more interesting things with video in general
  like you could build a judder-minimizing player by implementing something similar to madVR's smooth motion frame blending (not interpolation)
  • AV1? and codecs
  • WebRTC on UWP (Universal Windows Platform)
  • low-latency HLS (John Batos)
  • Device capability discovery (Will Law)
  • Chunk encoded-chunk transferred CMAF (Will Law)
  • Streaming ingest protocols (Josh Allmann)
  • Non-real-time encoding: transcoding and rendering effects in the browser... We'd like to be able to do in-browser editing of clips, from trimming / overlaying clips up to full editing or rendering of animations saved to a video. Today you have to either render such things on the server side, or render to a canvas and encode in real time using WebRTC-oriented interfaces (performs badly at high resolutions) or encode via WebAssembly?, which is much slower than native code or HW-accelerated encoding. (Brion Vibber)
  • Subjective video quality assessment (Chris Kennedy) Open source tooling https://github.com/crunchyroll/decima from Ellation built for Crunchyroll / VRV encoding research.
  • archive.org fun: AI/ML TV face recognition; OCR of TV chyrons; fully automatic websites via GitLab? "Auto DevOps?" using Kubernetes/Docker (all on premise / self cloud) (Tracey Jaquith)