Support for Audio File Transcription

What’s Coming

Upload audio files directly in Assistant or Vault and Harvey will automatically generate a transcript you can read, edit, and query.

Feature Highlights

Upload an audio file to Harvey, either directly in an Assistant conversation, or to Vault:

1. Harvey automatically transcribes the file in the background. This may take a few minutes depending on the length of the recording.
2. The original audio file is replaced by a Word (.docx) transcript. The transcript includes speaker labels (e.g. "Speaker A", "Speaker B") and timestamps at the start of each speaker's turn.
3. The transcript is ready to query in Assistant or Vault, just like any other uploaded document.

Please note: the original audio file is not stored in Harvey after transcription is complete. Only the generated transcript is saved. This applies to both Assistant and Vault. If you need to retain the original audio file for your records, make sure it is saved separately before uploading.

Supported file types and limits

Supported file types: M4A, MP3, WAV, WebM, FLAC, OGG

Supported Product Areas: Assistant, Vault. Workflow Builder will come later in the quarter.

File size limit: 500MB for assistant and 4GB for vault, and up to 2-hr of audio for each file type.

User flow

1. Upload an audio file to Harvey, either directly in an Assistant conversation, or to Vault (bulk upload of files is supported up to the file limits in each product area).
2. Harvey automatically transcribes the file in the background. This may take a few minutes depending on the length of the recording.
3. The original audio file is replaced by a Word (.docx) transcript. The transcript includes speaker labels (e.g. "Speaker A", "Speaker B") and timestamps at the start of each speaker's turn.
4. The transcript is ready to query in Assistant or Vault, just like any other uploaded document.

FAQs

Q: What security mechanisms are in place for audio file transcription?

All audio file inputs are transmitted using the existing security controls as outlined in Harvey's security documentation. There are no differences in encryption in transit or at rest.

Q: Where does transcription and data processing occur? Are records stored?

Transcription is processed in the same region as your Harvey workspace:

• app.harvey.ai — US data processing
• eu.app.harvey.ai — EU data processing
• au.app.harvey.ai — AU data processing (Azure only; ElevenLabs is not used)

Audio content is processed to generate the transcript and is not stored afterwards. Harvey, ElevenLabs, and Azure do not retain any audio recordings.

Q: How will this feature be access controlled?

Admins will be able to opt out for their workspace using a permission toggle. By default, the feature is on for all users in eligible workspaces.

Voice/audio files are treated the same way all other file inputs are treated by Harvey and its subprocessors. To learn more about Harvey's baseline data handling protections, see Your Data, Your Control: How Harvey Manages Customer Data.