Understanding HLS Latency: Causes and Solutions

HLS latency (the delay between a live event happening and a viewer seeing it) comes from several sources:

1. Encoding delay — Time to encode the video signal into the target format (typically 1-3 seconds)
2. Segment duration — The player must wait for a complete segment to be available. Typical 6-second segments add 6 seconds of delay
3. Segment availability — The server waits until the next segment is fully written before making it available
4. Playlist reload — The player periodically checks for updated playlists (every 3-6 seconds)
5. Buffer — Players maintain a buffer (typically 3 segment durations) for smooth playback

Combined, these factors result in a typical end-to-end latency of 15-30 seconds.

Apple introduced Low-Latency HLS (LL-HLS) as an extension to the HLS protocol. It reduces latency to 2-5 seconds through:

• Partial segments — Segments are made available before they're complete, as smaller "partial" segments
• HTTP/2 push — Server pushes updates to the player proactively
• Blocking playlist reload — The player requests the next playlist and the server holds the response until new content is available
• Reduced buffering — With lower segment duration and push, the player needs less buffer

• Reduce segment duration from 6s to 2-4s (increases server load but lowers latency)
• Use smaller GOP (Group of Pictures) sizes, matching segment duration
• Enable HTTP/2 on your streaming server for multiplexed requests
• Use a CDN that supports LL-HLS (most major CDNs do)
• Configure your player for lower buffer target (faster start but more rebuffering risk)

Lowering latency comes with trade-offs: shorter segments mean more files to manage, higher server load, and potentially more buffering events. Shorter GOP sizes reduce compression efficiency, increasing bitrate requirements for the same quality. For most content (VOD, pre-recorded shows), higher latency is perfectly acceptable. For live sports, auctions, or interactive streams, invest in LL-HLS or consider WebRTC for sub-second latency needs.