Notes from session:

1. Opus

  • High-level summary:
    • Low delay / relatime communication / collaboration
  • What is driving adoption?
    • Got strong consensus in Vancouver IETF meeting for Opus to be Mandatory To Implement in WebRTC (currently being confirmed on the mailing list)
  • Brief discussion of patent strategy:
    • Risks of having a very limited knob (Xiph-owned patents) to create a very complicated effect (providing safety for implementers).
    • How do you avoid patents?
      • The problems are the ones you don't know about
      • Doing things in the open, in standards bodies, has certain legal advantages that we didn't have for prior codecs

2. Daala / VPnext?

  • Goals
    • Have a royalty-free competitor to HEVC
    • Support the Web use cases: WebRTC, streaming, etc.
      • Low resolution performance important
      • Loss robustness important
      • Rate control adaptation/control important
  • Daala
    • Testbed for codec experiments, not a real codec yet
    • Trying to use lapped transforms to limit blocking artifacts, get better coding gain (textures, etc.)
    • To do this, need to solve a bunch of problems:
      • Intra prediction
      • Motion compensation
      • Etc.
    • Daala is the crazy experimental stuff that potentially has a high payoff, but may not work out.
  • VPnext?
    • Google's project to provide incremental improvements to VP8?
    • Each addition / modification may give only 1-2% gain, but 20 or 30 of those modifications start to add up
    • Already showing 25-30% gains (better gains at HD resolutions)
  • Standardization: Trying to form a working group in the IETF

3. transOgg

  • Rev of the Ogg container format to address lots of things we've learned since it was designed and much of the feedback we've received
  • Early design notes:
    • There are lots of points where there's no consensus
    • Please drive it