Skip to main content

​
What transcripts do

The Transcripts page (“Turn audio into speaker-labeled transcripts you can use in chats and projects.”) lets you create diarized transcripts in two ways:
  • Upload audio — upload an mp3, wav, m4a, mp4, or mpeg file and run it through a WhisperX transcription pipeline with pyannote speaker diarization.
  • Record a meeting — send a bot to a Zoom, Google Meet, or Teams URL, record the session, and receive an async transcript when the meeting ends.
Each transcript goes through a review step where you name the speakers and confirm the text before saving it to Knowledge.
Both flows are feature-gated. Upload audio requires a Whisper transcription backend configured by your administrator. Record a meeting requires a meeting bot integration. If only one is configured, you see only that flow. If neither is configured, the Transcripts page is not available.

​
The transcripts page

Open Transcripts from the app sidebar. The page is organized into three groups: Use the Search transcripts… field to filter across all three groups by title.

​
Upload audio

1

Open the composer

Click New transcription on the Transcripts page. If both audio upload and meeting recording are enabled, select Upload audio from the mode toggle.
2

Choose your file

Drag a file onto the drop zone or click to browse. Accepted formats: mp3, wav, m4a, mp4, mpeg.
3

Set options

  • Title — give the transcript a name (defaults to the filename).
  • Language — choose a language or leave on Auto-detect language.
  • Speakers — enter the number of speakers or leave on Auto-detect speakers.
  • Project — optionally attach the transcript to a project.
  • Sharing — set who can see the saved transcript (Private, shared with specific users, or shared with roles).
4

Start

Click Start. The status row appears in the Processing group (“Uploading…” then “Transcribing {length} of audio…”). When processing finishes, the row moves to Needs review.

​
WhisperX diarization

The transcription pipeline uses WhisperX with pyannote speaker diarization. The output is a set of time-stamped segments, each labelled with an auto-generated speaker label such as SPEAKER_00 or SPEAKER_01. You rename these labels in the review step.

​
Monthly recording usage

When a meeting bot integration is configured and a monthly recording-time cap is set, IMP shows a usage bar: Recording time this month with the format “{used} of {limit} used ({percent}%)”. When the cap is reached, the bar shows “Monthly recording limit reached — new meeting recordings are blocked until next month.” Whisper audio uploads are not counted against this cap.

​
Transcript library

Saved transcripts also appear in the transcript library at Data → Transcriptions (/data/transcriptions). The Transcript library link in the Saved group header takes you there directly. Re-opening a saved transcript from the Transcripts page lets you rename speakers; saving updates the existing library item.

​
Next steps

Record a meeting

Send a bot to join and record a live meeting.

Review and save

Name speakers, follow the audio timeline, and save to Knowledge.