Audio Transcribe

Overview

The Audio Transcribe node converts an audio input into a plain-text transcript using a speech-to-text provider. It accepts an audio file URL or binary stream on its input port and emits the resulting transcript string on its output port. If transcription fails — due to an unsupported format, provider error, or credential issue — execution is routed to the error port rather than halting the workflow. A valid provider credential must be configured before the node can run.

Configuration

Field	Type	Required	Description
provider	enum	Yes	Speech-to-text provider to use. Supported values: openai_whisper, google_speech, assemblyai, deepgram.
language	string	No	BCP-47 language code to hint the transcription engine (e.g. en-US, tr-TR). Leave empty to enable auto-detection.
outputFormat	enum	No	Shape of the transcript on the output port. plain_text returns a single string; timestamped returns an array of {start, end, text} segment objects. Defaults to plain_text.
credentialId	string	Yes	ID of the stored provider credential to authenticate the transcription request.
maxAudioDurationSeconds	number	No	Hard cap on input audio length in seconds. Inputs exceeding this limit are routed to the error port instead of being transcribed. Defaults to 600.

Inputs

Port	Type	Description
audio	AudioInput	The audio to transcribe. Accepts a publicly accessible URL string, a base64-encoded data URI, or a binary buffer reference produced by an upstream node such as File Read or HTTP Request.

Outputs

Port	Type	Description
transcript	string or transcript segments	The transcribed text. A plain string when outputFormat is plain_text; an array of {start: number, end: number, text: string} objects when outputFormat is timestamped.
error	NodeError	Emitted when transcription fails. Contains code, message, and provider-specific detail. Downstream nodes can inspect this to implement retry or fallback logic.

Example

{
  "nodeType": "audio_transcribe",
  "config": {
    "provider": "openai_whisper",
    "credentialId": "cred_whisper_prod",
    "language": "en-US",
    "outputFormat": "plain_text"
  }
}