Dictate code, commands, and half-finished thoughts on your Mac

Natter transcribes speech locally, fixes technical terms and sentence boundaries, then types into the app you were already using.

Download for macOS

v0.2.0 · 35.3 MB · Apple silicon, macOS 15+

Star on GitHub
0 stars · Open source
Verify the download
Claude Code~/dev/sites/ian.is
> ▍
Agent Listening

Tab: switch mode · tap ⌥: stop✕ Cancel

Double-tap ⌥ in any app and just talk. Natter types where your cursor was.

What Natter understands that normal Mac dictation misses

Talk into the current text field

Double-tap Right Option to start. Natter shows a live transcript, then types the result where your cursor already was.

Keep code and commands intact

Technical cleanup handles GitHub, SwiftPM, flags, paths, version numbers, commands, and names from your own dictionary.

Pick the rewrite level

Raw, Agent, Clean, Email, and Article modes range from direct transcription to local restructuring after you stop.

Nothing you say gets lost

If focus moves, keyboard events fail, or a rewrite is rejected, Natter keeps a local recovery record and copies the intended text.

No cloud transcription

Audio and transcripts stay on the Mac. There is no account, telemetry, localhost service, or Ollama dependency.

Personal rules are plain files

Preferred spellings and writing rules live as editable Markdown, including corrections you can teach by voice.

Your voice and transcript stay on the machine

Models install only when chosen
The required speech model and optional writing models download after you approve them and remain under Application Support.
Permissions have narrow jobs
Microphone captures speech, Accessibility finds the focused field, and Input Monitoring handles the global shortcut.
Signed and notarised by Apple
Public builds are signed with a Developer ID and notarised, so macOS can verify nothing changed since the build.
No account, no telemetry
There is no sign-up, usage tracking, localhost service, or Ollama dependency. The app is complete on its own.

Common questions about Natter

Does Natter send audio to the cloud?

No. Speech recognition and optional writing cleanup run on the Mac. Audio and transcripts are not uploaded.

Which Macs can run it?

Natter supports Apple silicon Macs running macOS 15 or later.

Do I need the large writing models?

No. The speech model is required. Raw and Clean modes do not need a writing model, and Agent has a deterministic fallback when its optional model is unavailable.

How do I start voice typing with Natter?

Double-tap Right Option while a text field is focused. Natter shows the live transcript and types the result back into that field when you stop.

Can Natter dictate terminal commands and file paths?

Yes. Its technical cleanup is built for commands, flags, paths, package names, version numbers, and preferred spellings from your personal dictionary.

What happens if the target app loses focus?

Natter keeps a local recovery record and copies the intended text when it cannot safely type into the original field.

Can I use raw transcription without rewriting?

Yes. Raw mode returns direct speech recognition. Clean, Agent, Email, and Article modes add different levels of local cleanup.

Say the messy version. Get useful text in the field you were already using.

Download for macOS

v0.2.0 · 35.3 MB · Apple silicon, macOS 15+

Star on GitHub
0 stars · Open source
Verify the download