Actions a_transcribe_stream

Live Transcription

Streams session audio through a real-time STT provider and emits transcripts.

What does it do?

Converts audio to text in real time, while the person is speaking.

How does it work?

Connects to phone or voice channels and transcribes the continuous audio stream without waiting for the user to finish.

Why use it?

Only used for phone calls (Voice Channel). Lets the bot respond lightning-fast in a spoken conversation.

Inputs

name	type
`audio`	stream

Outputs

name	type
`transcript`	string
`language`	string
`segments`	array
`provider`	string
`duration_ms`	number
`spawned`	boolean

Branches

Fixed branches

out — Output

Configuration

language string required default: "es"

model string optional

mode "await_final" | "spawn" required default: "await_final"

timeout_ms integer required default: 60000

emit_partials boolean required default: true

memory_key string optional

output_field string required default: "transcript"

sample_rate integer optional

codec "pcm16" | "opus" | "webm-opus" | "mulaw" | "flac" required default: "pcm16"

spawn_per_final boolean required default: false

audio_session_suffix string optional

speaker_role string optional

diarize boolean required default: false

max_speakers integer optional

Default configuration

{
  "language": "es",
  "mode": "await_final",
  "timeout_ms": 60000,
  "emit_partials": true,
  "output_field": "transcript",
  "codec": "pcm16",
  "sample_rate": 16000,
  "spawn_per_final": false
}

Emitted events

node.token
Streaming transcription tokens.
transcript.final
Final transcript for a segment.
transcript.done
Transcription completed.
transcript.error
Transcription error.
transcript.spawn_error
Failed to start transcription session.