AnchorNet: Bridging Live and Collaborative Streaming with a Unified Architecture
summary
AnchorNet implements a unified stream path with the RTC SFU as the single CDN publisher and server-side mixing, solving mode switches between single-host and collaborative streams. It adds audio sample-level splicing (duplicate tail samples, add priming to align frames) to handle encoder delays and codec hops (ASA <-> Opus transcoding). A four-week AB test with tens of millions daily shows video rebuffering drops of ~60–80%, engagement up to 3.83% and viewer watch time increases near 4%, so the system demonstrably works.
AI-written summary · about
chapters (8)
- 0:00AnchorNet introduction
- 1:21Live vs collaborative streaming
- 3:08Server-side stream mixing
- 6:41DualNet to AnchorNet
- 9:13End-to-end paths in AnchorNet
- 10:59Audio splicing techniques
- 11:19AnchorNet performance gains
- 12:44Future AnchorNet optimizations
related talks
17mLive Video Streaming as "tail -f"
19mLow Latency Live from a Different Vantage Point
11mFree ABR Renditions for User Generated Content Platforms
34mLightning Talks
18mDelivering Reliable Live Streaming Over Unreliable Backbone Networks
22mProviding better video experiences for the next billion users.
41m8K Codec Low-Latency Livestreaming, and AI Streaming
24mDynamic content insertion into live streams across the 3 main streaming formats