Voxdub MCP server

What the MCP server does

Who can use it

The MCP server is available with an active Voxdub plan or during a trial. The trial lasts 7 days and needs a card. After that, the plan costs 12.99 EUR per month or 99 EUR per year.

Get an access token

1. Sign in to Voxdub and open Settings → Developer.
2. Create a personal access token and give it a name (for example "Claude Code").
3. Copy the token at once. Voxdub shows it only one time.
4. Keep the token secret, like a password. If you lose it, revoke it and create a new one.

https://app-3.voxdub.com/settings?tab=developer

Set up your agent

The server uses the Streamable HTTP transport at https://mcp.voxdub.com/mcp. Each request sends the token in the Authorization: Bearer <token> header. Replace <token> with your token.

claude mcp add --transport http voxdub https://mcp.voxdub.com/mcp \
  --header "Authorization: Bearer <token>"

# ~/.codex/config.toml
[mcp_servers.voxdub]
url = "https://mcp.voxdub.com/mcp"
bearer_token_env_var = "VOXDUB_TOKEN"

export VOXDUB_TOKEN="<token>"

{
  "mcpServers": {
    "voxdub": {
      "url": "https://mcp.voxdub.com/mcp",
      "headers": {
        "Authorization": "Bearer ${env:VOXDUB_TOKEN}"
      }
    }
  }
}

OAuth sign-in (claude.ai and ChatGPT connectors) is not available yet. For now, use a client that accepts an Authorization header.

Step-by-step example

You can simply describe the result to your agent. For example:

"With Voxdub, create a rythmo band for scene.mp4 (French, 2 characters), fix the typos, apply the studio style, and export in 9:16."

1. 1

   Check the account

   get_account shows your email, your plan status, and your limits. The agent calls it first.

2. 2

   Upload the video

   create_upload takes the file name, its exact size in bytes, its type (video/mp4, video/quicktime, or video/x-matroska), and the spoken language. It returns a put_url that is valid for 15 minutes. The agent sends the file to that URL, then calls complete_upload, which returns the project_id.

   curl -X PUT -H "Content-Type: video/mp4" \
     --upload-file scene.mp4 "<put_url>"

3. 3

   Transcribe

   transcribe starts automatic transcription. The language is detected automatically, and speaker detection creates one character per voice. You can set min_speakers and max_speakers (1 to 20). The agent follows the job with wait_for_job until finished is true.

4. 4

   Fix the lines

   get_project gives the source band id and the characters. list_lines reads the lines (pages of up to 200). update_lines fixes up to 100 lines at once: text, start_ms, end_ms, and character. assign_character and update_character set the characters, their names, and their colors.

5. 5

   Set the style

   set_band_style applies a preset (standard, large, studio, comic, or reset) or exact settings: font, size, color, scroll speed, lane count, and band height. translate_band can also create bands in 1 to 5 more languages.

6. 6

   Export

   export_video renders the video with the rythmo band burned in, in 16:9, 9:16, or square, at 720p or 1080p, at 24, 25, 30, or 60 frames per second. The remove_voice option removes the original voices.

7. 7

   Download

   When wait_for_job shows the export is done, get_download_url gives a download link that is valid for 1 hour.

Tool reference

Limits

• 1 export running and 2 open jobs at most at the same time, per account.
• Videos up to 5 GB, in mp4, mov, or mkv.
• Exports up to 1080p.
• File upload only: import from a web URL is not available yet.

Troubleshooting