Overview

Nodes transform media between sources and sinks.

Available types

Type	Output	Description
video_mixer	Video	Composite layout (grid, speaker, custom)
audio_mixer	Audio	Mix multiple audio inputs
audio_resample	Audio	Convert sample rate/channels per track (n:n)
video_encoder	Encoded video	Encode raw video access units
audio_encoder	Encoded audio	Encode raw PCM access units
asr	Captions	Speech-to-text subtitles
translate	Audio + data	Managed speech-to-speech translation
voice_agent	Audio	AI voice agent (speech-to-speech)

Shape

{
  "name": "mix_video",
  "type": "video_mixer",
  "inputs": ["room_src"],
  "config": { }
}

An optional select on an object-form inputs entry filters which upstream tracks feed the node on that edge.

Structured output

asr, translate, and voice_agent always emit structured text alongside (or instead of) media. video_mixer and audio_mixer can emit layout and speaker level metadata when asked. See Data events for the payloads and how each sink delivers them.

Pricing

Per-node rates in Node pricing.