Works in every app
Messages, notes, documents, AI prompts. Insert automatically, by clipboard paste or by typing — with an optional clipboard fallback.
v0.1.8Runs 100% on your computer · macOS, Windows & Linux
Speak. Release.
Keep writing.
Hold a shortcut, say what you mean, let go. Plume types your words into whatever app you're in — privately, with no account and no audio upload.
Hold Ctrl Space to talk
Notes — Product sync
Product sync
Ship the onboarding flow behind a flag first, then open it to everyone once the copy is final.
Let's also add a short video on the download page so people see how the shortcut works before they install it
- 01
Hold the shortcut
Wherever your cursor is — a message, a doc, a prompt. Plume starts listening.
- 02ListeningEsc
Say it naturally
Talk the way you think. Prefer hands-free? Press Ctrl + Shift + Space to start and stop.
- 03Inserted
Release. It's typed.
Your text lands right where you were writing. Changed your mind mid-sentence? Esc cancels.
Features
Small app.
Out of your way.
Fully local
Speech is transcribed on your machine. Internet is only used to download models and check for updates.
Languages
Whisper models understand 99 languages. Let Plume detect it, or pin French or English.
Never lose a take
Local history with retention controls. Copy, delete, or recover your eight latest recordings.
Lives in your menu bar
Quiet in the menu bar or system tray. Light, dark, or follows your system.
Native & fast
Built in Rust with GPUI. No Electron, no browser tab hiding in the background.
Models
Pick your trade-off.
Download a model once from inside the app. Switch anytime. Everything runs offline after that.
| Model | Download | Languages | Best for |
|---|---|---|---|
| Nemotron 3.5 Compact | 793 MB | 35 | Fast responses |
| Whisper Base | 142 MB | 99 | Smallest download |
| Whisper Small | 466 MB | 99 | Balanced |
| Whisper Large-v3-Turbo | 1.5 GB | 99 | Highest accuracy |
Privacy
Your voice stays on your computer.
No account to create, nothing to sign into, no audio sent anywhere. What you say is between you and your machine.
Transcribing on-device
- 0
- accounts required
- 0
- bytes of audio uploaded
- Local
- history, stored as 16 kHz WAV on disk
- 1 click
- to clear history and its audio
Start talking to your keyboard.
Grab the installer for your system. Models download inside the app.
Questions
Does my audio ever leave my computer?
No. Transcription runs locally. Plume only goes online to download speech models and to check GitHub Releases for updates.
Why does my system show a security warning?
Installers aren't publisher-signed or notarized yet, so macOS and Windows may warn you on first launch. The full source is on GitHub if you'd rather build it yourself.
Can I change the shortcuts?
Yes, in Settings → Dictation. By default: hold Ctrl + Space to talk, Ctrl + Shift + Space for hands-free, Esc to cancel.
Which model should I pick?
Whisper Small is a good start. Go Large-v3-Turbo for the best accuracy, Nemotron for speed, or Base if disk space is tight.
Does it work on Wayland?
Not yet. Linux builds currently need X11 and a system tray. Wayland support is planned.