Zerm

Open-source desktop dictation

Zerm turns speech into clean clipboard text.

Tap a hotkey, speak, and paste. Whisper transcribes locally, Ollama can polish locally, Zerm can set it up for you, and the workflow stays private by default.

Download latest GitHub Releases View source
latest release MIT licensed No accounts or telemetry
01 Record

Right Option on macOS, Ctrl+Shift+Space on Windows and Linux.

02 Transcribe

Whisper runs locally with a bundled model.

03 Paste

The result lands in your clipboard, with optional native auto-paste on macOS.

Local first Audio and transcripts stay on the machine, with guided local setup.
Developer ready Agent, chat, pro, and raw transcript modes.
Auditable Tauri 2, Rust, TypeScript, diagnostics, and MIT license.

Native writing layer

Built around the OS behaviors that make dictation reliable.

Zerm is moving beyond transcription plus clipboard copy. The app now tracks setup diagnostics, insertion state, Accessibility trust, app signing, and the floating recorder HUD as first-class product surfaces.

01 macOS auto-paste

Uses Accessibility insertion with keyboard-event fallback, visible permission diagnostics, and signed prerelease checks for stable trust.

02 Full-screen recorder pill

The recorder overlay is hardened for full-screen Spaces, display changes, and explicit copied, pasted, permission, and failure states.

03 Cross-platform plan

Windows, Linux X11, and Linux Wayland have explicit strategy boundaries so unsupported paste paths are not presented as working.

Read native verification checklist

Downloads

Install a build or compile from source.

The cards update from GitHub Releases when a release is available. Current detected release: checking....

macOS Apple Silicon .dmg M1 and newer macOS Intel .dmg x86_64 Windows .msi / .exe Windows 10 and 11 Linux .AppImage Portable package Debian / Ubuntu .deb Install with sudo apt install ./Zerm.deb Source pnpm tauri dev Build locally

Privacy model

No hosted transcription. No analytics. History off by default.

Zerm records through the local microphone API, transcribes with a local Whisper model, and only sends text to your local Ollama service when a rewrite mode needs it. Zerm installs the official Ollama app when needed, while still letting developers explicitly opt in to an existing Homebrew or custom Ollama service.

  • Dictation history is opt-in.
  • Clearing history removes the backup state file too.
  • Setup downloads are bounded and verified where possible.
  • Existing Linux Ollama listeners are treated as unverified unless explicitly allowed.
  • macOS auto-paste uses Accessibility insertion with native key-event fallback.
  • Stable macOS and Windows releases require signing secrets before publishing; Linux artifacts ship with SHA-256 checksums instead of signing.
  • Prerelease macOS alpha builds are signed and notarized when release secrets are available.

Open source

Read the code, file issues, send patches.

The README includes architecture notes, platform setup, release process, and verification commands for local contributors.

Open repository