Skip to main content
Updated Sep 23, 2026

Transcribe Audio

Use this action when you have an audio file in your company storage and want a transcript produced from it.

Best for​

  • Transcribing a file a workflow just created, such as the output of Upload to storage or Download to storage
  • Turning voice notes or call recordings into text for later steps
  • Producing subtitle files (srt, vtt) alongside the plain transcript

To transcribe an audio message that is already in a conversation, use Execute function on document with the message's own transcription function instead — it is built for that case.

Main fields​

FieldWhat it does
FileThe audio file to transcribe, usually step(0).file from an earlier step
Bucket and Full pathAn alternative to File when you want to point at the object directly
Transcription profileWhich transcription settings to use. Leave it empty to use your company default
LanguageOverrides the profile's language, for example pt-BR
FormatsWhich outputs to produce: txt, srt, vtt, json. All four by default
PurposeOptional label for what this step is for, shown in the logs

What later steps can use​

ValueWhat it is
step(N).jobIdThe transcription job's ID
step(N).jobStatusThe job's status when it was queued — normally pending
step(N).enginePathHow the audio will be processed
step(N).providerTypeWhich transcription provider will run it
step(N).diarizedWhether the transcript will label who is speaking

This action queues the job — it does not wait for it​

The step hands back a job ID and finishes immediately. Real transcription takes minutes, so it runs in the background.

That means the transcript is not available to the next step in the same run. To get the result, arm a monitor in the next step: Create Monitor watching transcription_jobs / step(N).jobId with the condition doc.status == 'done' || doc.status == 'failed' || doc.status == 'canceled'. When the job finishes — seconds or hours later — the agent is woken in the same conversation and can read the transcript and deliver it.

You can also open the transcriptions panel in the app, or look the job up by its ID later; a notification is sent when the job finishes.

Tips​

  • You can queue up to 10 transcription jobs at a time. Beyond that the step fails until earlier jobs finish.
  • Transcription counts toward your plan's usage. Profiles that use your own provider keys are tracked but not billed.
  • Read the job's outcome with step(N).jobStatus, not step(N).status — status is reserved for HTTP status codes and will not hold the job's state.