Node pricing
Every node rate:
| Node | USD/min | Billed on |
|---|---|---|
video_mixer | $0.001 | While compositing, any input count |
audio_mixer | $0.00025 | While mixing |
audio_resample | $0.00005 | Per active track-minute |
asr | $0.012 | VAD-selected speech only |
translate | $0.04 | Provider session time |
voice_agent | $0.10 | Provider session time |
Mixers
Section titled “Mixers”Mixers are a flat rate per node-minute regardless of how many inputs they
composite or mix, billed only while frames are actually processed —
video_mixer meters wall-clock videoMillis, not the nominal canvas.fps.
Per-track audio processing
Section titled “Per-track audio processing”audio_resample bills per active track-minute, from the sum of processed
audio duration across all tracks. It is n:n, so simultaneous tracks are billed
independently. For example, three tracks processed for 60 minutes produce 180
active track-minutes. inputCount is not applied as a second multiplier.
Managed AI nodes
Section titled “Managed AI nodes”ASR, Translate, and Voice Agent use AVFlow-managed services. Their listed rates include the underlying AI service and AVFlow orchestration, so there is no separate vendor bill to reconcile.
Billed per minute of VAD-selected audio actually sent to the recognition service
(asrProcessedMillis), so silence removed before transmission costs nothing. The
processed duration does include the short preroll and hangover retained around
each speech segment, so words are not clipped.
Translate
Section titled “Translate”Billed while the speech-translation provider session is connected, measured by
activeConnectionSeconds. The managed rate includes OpenAI or Gemini
translation and orchestration. Diagnostic input/output audio durations are
reported separately and are not what you are charged on.
Voice agent
Section titled “Voice agent”Billed while the voice-agent provider session is connected, also by
activeConnectionSeconds. The rate includes the voice service itself.