Transcribing Audio Files In Harvey
Upload audio files directly in Assistant or Vault and Harvey automatically generates a Word transcript you can read, edit, and query.
Last updated: Jul 16, 2026
Overview
When you upload an audio file directly in an Assistant conversation or to a Vault, Harvey automatically converts it into a Word (.docx) transcript with speaker labels and timestamps.
The transcript is fully searchable, editable, and queryable, so you can turn a raw recording from a deposition, client interview, hearing, or call into usable text without leaving Harvey.
Each line in the transcript will follow this format:
[00:01:32] Speaker 1: Hello there.
[00:01:35] Speaker 2: Hello, nice to see you.
Before You Get Started
- Supported file types: M4A, MP3, WAV, WebM, FLAC, OGG
- File length: Files over 2 hours long will not be transcribed and will show an error.
- Language: Language is auto-detected. If a recording is not in a supported language or contains significant background noise, accuracy may be lower.
Important: The original audio file is not stored in Harvey after transcription. Only the generated transcript is saved. If you need to keep the original recording, save it separately before uploading.
Using the Audio File Transcription Feature
- Upload an audio file directly in an Assistant conversation or to Vault.
- Wait for Harvey to transcribe the file and convert it into a Word (.docx) file. This may take a few minutes depending on the length of the recording.
- Open the Word (.docx) transcript, which has replaced the original audio file.
- Review and, if needed, edit the transcript. See Editing a Transcript below.
- Query the transcript in Assistant or Vault.
Editing a Transcript
There are two ways to correct a transcript, such as renaming a speaker label or fixing an inaccuracy:
- Ask Assistant. In an Assistant thread, request the change in plain language. For example, "Rename Speaker 2 to 'Opposing Counsel' throughout the transcript."
- Edit in the Word editor side panel. Open the transcript and make corrections directly in the document, the same way you would edit any Word file in Harvey.
Tip: Correct any speaker labels or inaccuracies before running queries, so your results reference the cleaned-up transcript.
Querying a Transcript
Once the transcript is generated, you can query it like any other document: ask for a summary, pull key admissions, extract action items, or run it as part of a larger review. In Vault, the transcript is available to query alongside the rest of your documents.
To learn more, see Getting Started with Assistant and Vault: Analyze Large Document Sets at Scale.
Admin Controls
Audio file transcription is on by default for all users in eligible workspaces. Workspace admins can disable it for the entire workspace using a permission toggle.
For more on managing workspace permissions, see Access Control & Permission Management.
To learn more about Harvey's baseline data handling protections, see Your Data, Your Control: How Harvey Manages Customer Data.