MCP server for audio and video
An MCP server for audio and video work
Add SpeakerSplit to your AI assistant, then ask for what you need in plain words: split this interview by speaker, give me subtitles for this video, take the vocals out of this song. The assistant runs the job on your SpeakerSplit account and brings back the transcript or the download links.
Connect your assistant ยท See plans
Connect your assistant
Every assistant asks for the same address: https://speakersplit.io/mcp. You sign in to your own SpeakerSplit account and press "Allow". There is no key to copy.
The address
https://speakersplit.io/mcpClaude (website or desktop app)
- In Claude, open Settings, then Connectors.
- Choose "Add custom connector".
- Paste the address https://speakersplit.io/mcp
- Choose "Connect".
- Sign in to SpeakerSplit and press "Allow".
Claude Code
Run this once in your terminal.
claude mcp add --transport http speakersplit https://speakersplit.io/mcpCursor
Add this to your mcp.json file.
{
"mcpServers": {
"speakersplit": {
"url": "https://speakersplit.io/mcp"
}
}
}ChatGPT
Turn on developer mode, then add a connector with the same address. Sign in to SpeakerSplit and press "Allow". In ChatGPT you can attach the recording to the chat instead of pasting a link.
The 14 tools
| Tool | What it does | Cost |
|---|---|---|
| split_speakers | Splits a recording into one track per speaker, with a transcript. | Credits |
| transcribe_audio | Writes a transcript with each speaker named. Ask for subtitles to get SRT or VTT. | Credits |
| remove_vocals | Separates the voice from the music: a vocals track and an instrumental track. | Credits |
| split_stems | Splits a song into vocals, bass, drums, piano, guitar and other. | Credits |
| enhance_audio | Removes background noise from speech. | Credits |
| add_captions | Burns captions into a video, in time with the speech. | Credits |
| remove_silence | Cuts pauses out of a recording, and um and uh in English. | Credits |
| translate_audio | Translates what is said into one of 21 languages, read aloud by an AI voice. | Credits |
| change_voice | Re-speaks a recording in a different AI voice. | Credits |
| get_job | Shows how far a job has got, and the download links when it is done. | Free |
| get_transcript | Returns the transcript with speaker names, as text, subtitles or JSON. | Free |
| list_jobs | Shows your most recent jobs and their results. | Free |
| get_account | Tells you how many credits you have left. | Free |
| create_upload | Gives a temporary upload link for a file on your computer. | Free |
Getting your file to the assistant
Paste a public link to the file. In ChatGPT, attach the file to the chat. For a file on your computer, create_upload gives a temporary upload link, which an assistant that can run commands, such as Claude Code, can send the file to.
What it costs
Each job uses credits from your SpeakerSplit account, at the same price as on the website. Checking a job, reading a transcript, listing your jobs, checking your credits and getting an upload link are free.
Common questions
What is an MCP server?
MCP (Model Context Protocol) is an open standard that lets AI assistants use outside tools. An MCP server is the service that offers those tools. The SpeakerSplit MCP server gives your assistant 14 audio and video tools.
Which AI assistants work with the SpeakerSplit MCP server?
Claude on the website and desktop app (as a custom connector), Claude Code, Cursor, and ChatGPT in developer mode.
Do I need an API key?
No. When you connect, you sign in to your SpeakerSplit account and press "Allow".
What does it cost to use?
Each job uses credits from your SpeakerSplit account, at the same price as on the website. Checking a job, reading a transcript, listing your jobs, checking your credits and getting an upload link are free.
Can I attach a file instead of pasting a link?
In ChatGPT, yes: the job tools accept a file attached to the chat. In other assistants, paste a public link, or use create_upload for a file on your computer.