Início / Conhecimento / Ao vivo
Understanding and Reducing HLS Live Latency
Este artigo está disponível em chinês e inglês; a tradução para português está a caminho.
Low latency is the heart of a good live experience. To improve it, you first need to know what the delay is made of.
Where latency comes from
- Capture and encoding: camera capture and encoder processing take time.
- Segmentation: the encoder must fill a complete segment before publishing, so segment length contributes directly.
- Distribution: upload to the CDN, plus CDN caching and refresh.
- Player buffer: the buffer kept for jitter resistance is usually the largest share.
End-to-end optimisation
- Use low-latency encoder settings on the ingest side (short GOP, zero buffering);
- Shorten segments from 6–10 seconds to 1–2 seconds;
- Lower the player's buffer target and use LL-HLS incremental loading;
- Pick CDN nodes close to your audience.
Don't fixate on one stage. The whole chain is like a barrel: the shortest stave decides the final latency. Measure each leg first, then optimise where it counts.