Local voice dictation for Linux
Talk to your computer.
It types what you say.
Wisper listens through a global hotkey, transcribes on your machine with Whisper, and drops the text right where your cursor is. No cloud required. No keystrokes sent anywhere.
- 100% on-device by default
- Wayland & X11
- MIT licensed
From voice to text in one breath
A short pipeline with no round trip to a server. You speak, the local model listens, your words land.
-
01
Speak
Hold the global hotkey (push-to-talk) or tap for toggle mode. Your mic records locally.
-
02
Transcribe
Whisper ONNX models turn speech into text on your machine. Leading and trailing silence is trimmed so pauses don't linger.
-
03
Insert
The text is typed at your cursor or copied to the clipboard, in any app, on Wayland or X11.
Everything you need to stop typing
Wisper does one job and stays out of your way. These are the controls you actually get.
-
Speak instead of type
A global hotkey captures your voice and drops the text where you are working. Hold to talk, or tap to toggle.
-
Runs on your device
Transcription uses local Whisper ONNX models. Nothing leaves your computer unless you point it at a cloud provider.
-
Pick your microphone
Choose a specific input device or leave it on the system default. Virtual sinks are filtered out so you don't pick the wrong one.
-
AI cleanup
An optional step reformats and polishes the transcript. Profiles include Email, Developer, Messaging, Formal, Clean-up, and your own custom instructions.
-
Your words, your way
Custom vocabulary turns shortcuts into proper terms. Say "gpt", get "GPT". Scan past dictations to find new words worth saving.
-
Look back
Searchable history lets you replay the recording, re-transcribe, or edit any past dictation. It also shows how much typing time you saved.
-
Live overlay
A small indicator follows your cursor while recording, showing a voice-reactive waveform. If transcription fails, it flashes a clear mark.
-
Out of the way
Lives in the system tray. Close the window and it keeps running, ready for the next hotkey. Unload the model from the tray when you don't need it.
-
Updates itself
Checks GitHub releases and installs new versions in-app, with signed auto-updates. You stay current without hunting for downloads.
-
Local or cloud, your call
Run everything on your device, or skip the model download and point Wisper at a cloud speech provider you configure (OpenAI, Groq, or your own). Switch any time from the Engine tab.
-
Starts with you
Launch at login and start hidden in the tray. First run walks you through the model download and the difference between push-to-talk and toggle.
Download for Linux
Wisper v3.0.0 — the site fetches the latest release when it loads. Auto-update is built in.
Loading the latest release…
All releases are signed. Prefer the AppImage if your distro isn't DEB based. View all releases →
First-run setup
Three steps and you're dictating. The hard part is picking a good microphone.
1 Pick an engine
On first launch, open Engine. Download a local speech model (about 1.4 GB), or skip it and connect a cloud provider like OpenAI or Groq. Nothing transcribes until one is active.
Engine tab · local or cloud
2 Paste backend
Wisper inserts text by simulating a paste at your cursor. On Wayland, install ydotool for prompt-free pasting:
sudo systemctl enable --now ydotoold
sudo usermod -aG input $USERwtype also works on compositors that support the Wayland virtual-keyboard protocol. enigo is built in but prompts for the RemoteDesktop portal once on Wayland.
Wayland · ydotool recommended
3 Pick your mic
In General settings, choose a specific input device or leave it on system default. Test the level before you rely on it.
General tab · input device
Questions people ask
The honest answers, including the Linux-specific rough edges.
Is my audio sent to the cloud?
No, not by default. Transcription runs on your machine with local Whisper ONNX models. Cloud providers are optional and only used if you add a key in Engine or Process settings.
Which Linux desktops are supported?
Any Linux with X11 or Wayland. Wisper uses a global hotkey that works the same in both. For pasting without a permission prompt on Wayland, install ydotool.
Why does it ask for "remote desktop" permission?
That prompt comes from the built-in paste tool (enigo) on Wayland, which routes input through the XDG Desktop Portal. Installing ydotool or wtype avoids it, because they inject input through dedicated channels below the compositor.
Do I need to download a model?
Only if you want to stay fully local. On first launch, open the Engine tab and download a speech model (about 1.4 GB). Prefer not to? Connect a cloud provider like OpenAI or Groq instead. Wisper can't transcribe until one engine is active, and you can cancel a download mid-way.
What is the cloud option?
In the Engine tab you can switch from the local model to a cloud speech provider (OpenAI, Groq, or a custom OpenAI-compatible endpoint) using a key you supply. The audio then leaves your machine to that provider. Everything else, including the AI cleanup step, works the same.
What does the AI cleanup actually change?
It re-formats and polishes the transcript through an OpenAI-compatible chat model you configure. It rewrites only what you said, never answers for you. Profiles cover email, developer, messaging, formal, and plain cleanup.
Is Wisper free and open source?
Yes. The code is MIT licensed and the repository is public on GitHub. There is no account, no telemetry, and no paid tier.