Voice typing for Mac.
Tiny macOS app that turns speech to text — fast, accurate, and private. Bring your own API key, connect to the best speech models, and pay only for what you use.
Hi Daniel,
Hope you're doing well and that this email finds you before the deadline does. I still need the assets from your team — I've attached the specs again for the fourth time, which I believe makes this a tradition now.
Adapts to your needs
A few small settings make Typist fit the way you actually work.
Choose your models
Switch between Whisper, GPT-4o Transcribe, Voxtral, and Scribe V2.
WhisperGPT-4o TranscribeVoxtral Mini TranscribeScribe V2Personal dictionary
Teach it the names, jargon, and acronyms it would otherwise mishear.
Sveltellama.cppClaude CodeAdd term…Per-app instructions
Different tone or formatting in Slack vs. email vs. code. Set once per app.
Slackcasual tone, lowercaseMailformal, full punctuationTerminalterse, technicalNotionstructured, headings
Integrates with other tools
Typist plays nicely with the rest of your setup — no walled garden, no proprietary formats.
AppleScript commands
Scriptable from Shortcuts, Raycast, Alfred, or anything that speaks AppleScript.
tell application "Typist"start dictationend tellLocal, readable history
Transcripts are plain files on your Mac — searchable in Spotlight, Alfred, or Raycast.
2026-05-11 09-14-22.txt2026-05-11 10-02-47.txt2026-05-10 17-38-05.txt2026-05-10 14-11-19.txtText-based config
Under the hood, settings are plain text files — diff, version, or edit them directly.
~/.typistdictionary.txtinstructions.mdinstructions/Mail.mdSlack.mdTelegram.md
No subscription
You pay the transcription provider per minute of audio — no Typist account, no per-seat pricing.
Bring your own key
Paste your provider API key into Typist and start dictating. Switch providers whenever you want.
No accounts
No sign-up, no login. Nothing to forget, nothing to cancel.
No middleman
Typist isn’t a server. Audio goes from your Mac to the provider you picked — nowhere else.
No overreach
Typist tries to take as little as possible — of your data, your permissions, and your machine.
No data collection
No telemetry. Transcripts, keywords, and settings live on your Mac. The provider sees only the audio you send.
Minimal permissions
Asks for the microphone and Accessibility (to paste into the focused app). Nothing else.
Small footprint
0.5 MB on disk. Low CPU, memory, and battery use.
Changelog
Here's a summary of the last 16 updates:
- 0.31.0 May 05 Update app icon and HUD design
- 0.30.0 May 04 Add option to copy transcripts to clipboard
- 0.29.0 Apr 16 Redesign HUD with backdrop blur and refined shape
- 0.28.0 Mar 18 Simplify transcript history
- 0.27.0 Mar 18 Simplify app updater
- 0.26.0 Mar 12 Add Recent Transcripts submenu for browsing past transcripts
- 0.25.0 Mar 08 Redesign recording HUD and settings UI
- 0.24.0 Mar 03 Add new transcription providers and Whisper model support
- 0.23.0 Feb 09 Make Shift Shift the default activation shortcut
- 0.22.0 Feb 04 Add Enter Enter shortcut to transcribe and send messages
- 0.21.0 Feb 02 Make "Refine speech" more aggressive at detecting implicit restarts
- 0.20.0 Jan 24 Add intent-based speech refinement for cleaner transcripts
- 0.19.0 Jan 16 Update app icon and add dictionary editor with sidebar tabs in settings
- 0.18.0 Jan 13 Improve paste detection for Cursor and other Electron apps
- 0.17.0 Jan 11 Show transcript overlay when paste fails
Let's keep in touch
Occasional updates on new features, supported models, and notable changes. No spam, unsubscribe anytime.