Sub-frame Latency
summary
A complete capture-to-playback stack encodes and decodes video 16–32 lines at a time so presentation happens at sub-frame granularity, giving latency measured in lines (a line ≈ 40 μs) instead of milliseconds. The pipeline uses chunked encoding over the network at about 200 Mbps, chunk-based decoding and the draw horizontal band callback to render slices as they arrive; it works with added jitter-buffering but needs x264 subframe encoding and better FFmpeg APIs to lower bitrate and CPU cost.
AI-written summary · about
chapters (6)
related talks
19mLow Latency Live from a Different Vantage Point
23m3 Second End to End Latency At Scale
11mS-Frame in AV1: Enabling better compression for low latency live streaming.
24mLive Video Transmuxing/Transcoding
44mLive Video in a Modern Web Environment
20mYour TV Is Eating Your Frames
16mSwift: Adaptive Video Streaming with Layered Neural Codecs
41m8K Codec Low-Latency Livestreaming, and AI Streaming