Streaming Audio — Tips

Tuning Latency & Segmentation

How to use chunk_size_sec and silence_threshold_sec to balance latency vs accuracy.

Settings guide

Use case	Recommended settings
Voice agent / IVR	`chunk_size_sec=0.5`, `silence_threshold_sec=0.3`
Live captions	`chunk_size_sec=1.0`, `silence_threshold_sec=0.5` (defaults)
Lecture / dictation	`chunk_size_sec=2.0`, `silence_threshold_sec=1.5`