YouTube live captions and stream latency

Use a compatible latency mode

YouTube’s LiveBroadcast API documentation states that Ultra low latency does not support closed captions. Use Normal latency for the StreamTranslate setup described here. Low-latency behavior should be verified on your own private broadcast before relying on it.

Video latency is not caption processing time

The delay from speaking in your room to hearing your voice on YouTube includes encoding, network delivery and player buffering. Caption processing also has recognition, translation and posting stages. Judge synchronization against the audio in the same player.

Change one setting at a time

1. Create a private test and leave languages and audio unchanged.
2. Record the Studio latency mode.
3. Compare a short phrase and a longer passage with the matching player audio.
4. Try the Advanced later preset only if captions consistently lead.
5. Repeat the same speech before drawing conclusions.

Avoid a misleading speed promise

There is no fixed end-to-end delay that applies to every stream. Speech pace, sentence boundaries, translation, reconnects and YouTube playback can all change when text appears. A stable alignment test is more useful than a single advertised millisecond figure.

Keep a timing comparison log

For each private test, write down the latency mode, caption language and Advanced preset. Use the same short passage, a similar connection and the same player position near the live edge. If you pause the player or seek backward, return to the live edge before comparing another run. A single unusually fast or slow line is not enough to establish a stable setting. Look for repeatable alignment across startup, continuous speech and recovery after an interruption.

Sources and setup limits

See YouTube’s official requirements for the underlying caption transport and latency modes. Actual behavior must also be checked on your own broadcast.

• YouTube live caption requirements
• YouTube LiveBroadcast API and latency limitations