Scribe · AI transcription for Claude

Transcribe meetings, calls and videos — right inside Claude.

Paste a link or drop a file. Scribe & Imagine turns up to 4 hours of audio or video into an accurate transcript with speaker labels and subtitles, and Claude does the rest: summaries, decisions, action items, quotes, follow-up emails.

https://scribeimagine.com/api/mcp

Add it in Claude → Settings → Connectors. Setup guide →

An audio waveform on a laptop splits into three coloured ribbons that become stacked lines of text, one colour per speaker.

Long recordings, fast

Cut at natural pauses and transcribed in parallel: a one-hour meeting comes back in about a minute, up to 4 hours per file.

Who said what

Speaker labels that stay consistent across the whole recording; rename Speaker A to Dana once it's clear.

Spelled right

Tell Claude who's in the meeting and what it's about — names, products and jargon are passed to the model so they come out right.

Any format, any link

MP3, M4A, WAV, MP4, MOV, WebM… from Google Drive, Dropbox, OneDrive, Zoom or Loom downloads, or a direct upload.

Subtitles

Timestamped transcripts with .srt and .vtt downloads for video editors and YouTube.

Pay for speech, not silence

Long pauses are trimmed before sending, and the same file is never billed twice.

Try asking Claude

  • “Transcribe our weekly sync with speaker labels — it’s Dana, Sam and Priya: <Drive link>. Then list decisions and action items with owners.”
  • “Make English subtitles for this interview: <Dropbox link>.”
  • “Transcribe this 3-hour workshop and write a one-page summary with timestamps for each topic.”

Questions

Can Claude transcribe audio?

Claude can't listen to audio on its own, but with the Scribe & Imagine connector it can: give it a link to a recording (or upload the file) and it transcribes it with OpenAI's speech models, then summarises, searches or quotes the transcript for you.

How long can a recording be?

Up to 4 hours and 450 MB, in any common audio or video format (MP3, M4A, WAV, MP4, MOV, WebM and more). Long recordings are split at pauses and transcribed in parallel, so an hour usually takes about a minute.

Does it tell me who said what?

Yes. Turn on speaker labels and each paragraph is attributed to Speaker A, B, C…, kept consistent across the whole recording. Then tell Claude who is who and it renames them.

Can I get subtitles?

Yes — ask for timestamps and you can download .srt and .vtt subtitle files for YouTube, Premiere, DaVinci Resolve or any player.

How much does it cost?

With your own OpenAI key, a one-hour meeting costs about $0.27 at OpenAI's list price and we charge nothing. With prepaid credits it's OpenAI's price plus a 20% service fee. Long silences are trimmed so you don't pay for them.

What happens to my recording?

It's deleted as soon as it has been transcribed. The transcript stays in your account until you delete it.

See pricing → · Image generation in Claude →

Guides