What is Latency?
Latency is the elapsed time between an event or request and its observable result. In live streaming, it commonly means the delay from capture through processing and delivery to viewer playback.
How Latency works
Media delay accumulates across clocked stages rather than at one server boundary. Capture exposure, encoder lookahead, segment production, origin and edge transfer, player buffering, decoding, and display synchronization each contribute. Glass-to-glass delay measures the full live path, while network round-trip time, startup delay, and command response time describe different phenomena. Reliable optimization timestamps the same event against synchronized clocks and separates steady-state delay from jitter and tail behavior.
Key facts
- Glass-to-glass measurement needs a common time reference at capture and display; subtracting unsynchronized device clocks can produce a convincing but invalid result.
- Encoder lookahead and player buffer are controllable latency reservoirs, while CDN distance and retransmissions add variable delay that averages can conceal.
- A player can start near the live edge and later drift behind it after stalls, so production monitoring should track live-edge distance throughout the session.
When Latency matters
Measure latency across capture, encoding, packaging, transport, buffering, and rendering to locate the dominant delay. Reducing buffers improves responsiveness but raises the risk of stalls on unstable networks.