Live Q&A 1 with Speakers
summary
At scale, the system prioritizes end-to-end encryption prevents server decoding so heavy work moves to lightweight on-device pipelines: a low‑compute classical signal processing baseline plus optional ML models that scale with available CPU. Large group calls surface CPU-driven audio glitches and echo leaks, so servers apply selective forwarding unit avoids server decode with server-side audio capping; server mixing (decode, pick top four, re‑encode) trades CPU for per‑hop loss recovery. Spatial audio is ramping up for metaverse and grid views, AV1 is in Chromium, and new in-field, privacy-safe metrics replace lab VMAF.
AI-written summary · about
chapters (16)
- 0:00Live Q&A 1
- 0:25Server vs edge processing
- 1:48Reliability in large calls
- 3:52Mixing in RTC systems
- 5:15Spatial audio trends
- 6:38H.264 vs AV1
- 8:24Codec quality measurement
- 9:27Video quality trends
- 10:52AV2 and AV1 in browsers
- 11:59Simulcast and codec switching
- 12:40RTC innovation obstacles
- 15:26Adoption vs security
- 17:12Spatial audio usability
- 19:20Audio chat in gaming
- 21:25Audio realism benchmarking
- 24:12Closing remarks
