HuggingFace Account Setup

Speaker diarization in WhisperX uses pyannote.audio models hosted on HuggingFace. These models require a free HuggingFace account and an access token. This guide walks you through creating an account, generating a token, and accepting the required model license terms.

Step 1: Create a HuggingFace Account

1. Go to https://huggingface.co
2. Click Sign Up in the top-right corner
3. Complete the registration with your email address
4. Verify your email when prompted

Step 2: Generate an Access Token

1. Sign in to HuggingFace
2. Go to Settings > Access Tokens: https://huggingface.co/settings/tokens
3. Click New token
4. Configure the token: - Name: whisperx (or any name you prefer) - Type: Fine-grained - Permissions: Read access to repos is sufficient
5. Click Generate a token
6. Copy the token and save it somewhere safe — you will need it during LymeScribe server setup

Warning: Treat this token like a password. Do not share it or commit it to source control.

Step 3: Accept Model License Terms

Current WhisperX (3.8+) uses the pyannote/speaker-diarization-community-1 pipeline for speaker diarization. You must accept its license terms before the weights can be downloaded.

1. Visit the speaker-diarization-community-1 model page: https://huggingface.co/pyannote/speaker-diarization-community-1 - Read the license terms (CC-BY-4.0 — commercial use allowed, attribution required) - Fill in the requested contact info and click Agree and access repository (or similar button)

Note: Older guides reference pyannote/speaker-diarization-3.1 and pyannote/segmentation-3.0. whisperx 3.8+ defaults to the community-1 pipeline and ignores --diarize_model 3.1, so accepting community-1 is what matters. Accepting the 3.1/segmentation pages as well does no harm if you previously did so.

How to Verify Access

After accepting terms, revisit the model page. If you see a Download or Files and versions tab (not a "Request access" or "Agree" button), you have been granted access. The LymeScribe diarization setup wizard also verifies this for you against your token.

Step 4: Save Your Token

You will need this token when configuring the LymeScribe server. During the setup wizard, you will be prompted to enter your HuggingFace token. Keep it accessible.

Troubleshooting

Token not working / 401 Unauthorized errors

• Verify that you accepted the license terms for the community-1 model in Step 3
• Check that the token has not been revoked — go to Access Tokens and confirm it is active
• Regenerate the token if needed and update your LymeScribe configuration

"Access denied" when downloading models

• You must be logged in to the same HuggingFace account that accepted the model terms
• The token must belong to the account that accepted the terms
• Try visiting the model pages again while signed in to confirm access is granted

Models download slowly or fail

• HuggingFace model downloads can be large (several hundred MB). Ensure you have a stable internet connection.
• If downloads time out, try again — the download resumes where it left off in most cases