Transcribe Audio and Video Through AssemblyAI
Through an AnythingMCP custom connector, your muse can turn a recording at a public link into a transcript with speaker labels, summaries, chapters, and subtitles, and answer questions about it.
What it does
AssemblyAI is a speech-to-text service: you give it a recording, it gives back the words. Connected this way, the muse gets 13 tools. It can start a transcript from a public link to an audio or video file, with speaker labels, language detection, and extras such as a summary and chapters. It can hide personal data in the text, and make a copy of the audio with that data beeped out. When the transcript is ready, the muse can read it as text, sentences, or paragraphs, each with times, and export subtitles as SRT or WebVTT. It can search the transcript for a phrase and tell you where it was said. It can also translate a finished transcript, put real names to the speakers, or answer a question about it, such as listing action items and their owners.
How to use it
- Start an AnythingMCP account and add the AssemblyAI connector.
- Make an AssemblyAI account, open API Keys, and copy a key. Enter it only where the connector asks, never in a chat. Choose the US or EU region it asks for.
- In Muse, open Settings, then Connectors, and add a custom connector with the MCP server address AnythingMCP gives you. Sign in and approve.
- Give the muse a public direct link to one recording and say what you want back, for example, speaker labels and a summary. For a long file, ask the cost first.
- Check the result against a short stretch of the recording, especially names, before sharing it.
What you need
- A Muse account that can add custom connectors
- An AnythingMCP account with the AssemblyAI connector added
- An AssemblyAI account and API key, with a recording at a public direct link
What to watch for
- Local files cannot be uploaded through this connector, and a page link does not work. The link must point straight at the file.
- Transcribing is paid by the hour of audio, with extras such as speaker labels costing more. New accounts get free credit, and failed transcripts are free.
- Transcripts made in the US region are not visible from the EU region. Pick one and stay with it.
- Names and unusual words can come back wrong. A name list helps, and a quick listen check still matters.
- Deleting a transcript is switched off at install, because it cannot be undone.