Your AI agents can transcribe, right on your Mac.
Claude Code, Codex, Cursor and other agents that run commands can turn a meeting, voice memo or video into text with SpeedOfTalk. Ask in plain words; the agent does the rest. Nothing is uploaded.
10 minutes of audio in about 3 secondsSpeakers labeledWord for word
Set it up once
- Install SpeedOfTalk.
- For Claude Code: open Settings → General → AI Agents and click Add to Claude Code.
- For other agents: give them the skill file, a short page of instructions they read when you ask for a transcript.
Then just ask
You ask your agent
Transcribe standup.m4a with speakers and save it as Markdown.
It runs
SpeedOfTalk --transcribe standup.m4a --speakers --format markdown > standup.md
What comes back
Real output from a short two-person recording, unedited. It took 1.3 seconds on an M3 MacBook, including loading the speech model.
# standup
**Speaker 1** · `00:00`
Morning, quick one before stand-up, did the invoices go out on Friday?
**Speaker 2** · `00:04`
They did, too bounced, so I'm resending those this afternoon.
**Speaker 1** · `00:09`
Great, let's move the design review to Thursday, and I'll send the agenda.
**Speaker 2** · `00:14`
Thursday works, I'll book the room.
Notice "too bounced" where the speaker said "two bounced": transcripts are word for word, with no rewording or AI clean-up, so you see exactly what was heard. Names you add to your dictionary are spelled your way.
Formats
- Text, plain and simple (the default)
- Timestamped text
- Markdown, as above
- SRT and VTT subtitles
Any audio or video file your Mac can play works: m4a, mp3, wav, aiff, mp4, mov and more. You can also drop files into SpeedOfTalk's Transcribe window yourself.
How fast
A 10-minute recording took about 3 seconds on an M3 MacBook Pro, or 10 seconds with speakers labeled. Speech recognition runs on your Mac's own chip, so there's no upload and no queue.
Questions
Which agents work?
Any agent that can run commands on your Mac: Claude Code, Codex, Cursor and others. The skill file tells them how.
Does it cost extra?
No. It's part of SpeedOfTalk: free during the 7-day trial, then included with your license.
Which languages?
The same 25 European languages as dictation, detected automatically. See the FAQ.