Speak.
The rest is text.

Simplevoice is a privacy-first, fully offline speech-to-text app for the desktop. Press a key, talk, and your words land in any app. No cloud, no account, no telemetry.

Download for desktop View on GitHub
Available for macOS Windows Linux
Watch demo
Demo

See Simplevoice in action.

From a single keystroke to finished text: a full dictation, running entirely on-device.

How it works

Three keys to finished text.

No setup ritual. No dashboard to learn. Speak, and it's written.

01

Press the hotkey

Hit the shortcut in any app. A floating overlay appears and starts listening, wherever you are.

Ctrl / Cmd + Shift + Space
02

Speak naturally

Talk at your own pace. Voice Activity Detection notices when you stop and transcribes on-device, automatically.

03

Text lands where you type

Your words are pasted straight into the active field, or copied to the clipboard with a second hotkey.

Features

Everything you need to dictate, nothing you don't.

A focused tool that does one job exceptionally well: capture your voice and turn it into finished text, wherever you work.

Record natively, stop automatically.

High-fidelity capture through the Rust-native CPAL library, wrapped in a sleek floating overlay. Voice Activity Detection ends the take and starts transcribing the moment you stop speaking.

0:07

Recognition, entirely on-device.

Four local engines run Whisper, Parakeet, Moonshine and more, completely offline. Hardware acceleration with Metal on macOS and Vulkan on Linux and Windows, plus a safe CPU fallback.

Whisper.cpp
Whisper GGML and GGUF, tiny to large
Metal, Vulkan
Candle
Whisper, Wav2Vec2 and CTC models
CPU, Metal
sherpa-ONNX
Parakeet TDT v3, Moonshine
CPU
NeMo
FastConformer-CTC, Conformer-RNN-T
Python sidecar

One hotkey, any app.

A global shortcut records from anywhere in the OS, and your dictation is typed straight into the active field. Auto-paste is pure Rust on Wayland, with native paths for macOS, Windows and X11.

The active text field
On my way, be there in ten.

Cloud, only if you ask.

Optional bring-your-own-key transcription. Pick a provider, paste a key, and switch back to local whenever you like.

OpenAI OpenRouter Google Gemini Custom endpoint
Keys live in the OS keyring
macOS Keychain, GNOME Keyring, Windows Credential Manager

Models, managed in-app.

Browse recommended models, download them with live progress, or import your own and drop them into the models directory.

parakeet-tdt-0.6b-v3.int8.onnx 306 of 478 MB

Your history stays home.

Every transcription, duration and word count lives in a local SQLite database, with 7-day, 30-day and all-time views. Nothing syncs, nothing uploads.

Last 7 days 4h 32m dictated
Privacy

Your voice never leaves your machine.

Simplevoice is private by architecture, not by policy. Recognition runs fully on-device, and nothing is sent anywhere unless you explicitly choose a cloud engine.

Offline by default

With a local engine, your audio is transcribed entirely on your device. No internet required.

  • Zero telemetry

    No analytics, no tracking, no phone-home. There is nothing to opt out of, because we collect nothing.

  • No account, ever

    No sign-up, no login, no profile. Install it and start talking. You stay anonymous.

  • Keys in the OS keyring

    If you opt into a cloud engine, API keys live only in your operating system's keyring, never on disk or in localStorage.

  • Open source, Apache-2.0

    Every line is inspectable and auditable on GitHub. Trust the code, don't just take our word for it.

Download

Say the word.

Free and open source. Installers for macOS, Windows and Linux, built and published on GitHub Releases.

Download for desktop View on GitHub

Version 0.1.0, Apache-2.0

macOS
Apple Siliconarm64Intelx86_64
Windows
Installerx86_64
Linux
AppImagex86_64.debDebian / Ubuntu.rpmFedora / RHEL
AURArch / Manjaro
$yay -S simplevoice-bin
View on AUR

Prefer to compile it yourself? Build from source

FAQ

Questions, answered.

Everything worth knowing before you start talking.

Is Simplevoice really free and open source?
Yes. Simplevoice is licensed under Apache-2.0 and the full source lives on GitHub. There are no paid tiers and no accounts.
Does it work completely offline?
Yes. With a local engine (Whisper.cpp, Candle, sherpa-ONNX with Parakeet, or NeMo) transcription runs entirely on your device, with no internet connection required.
Do you collect any data or telemetry?
No. Zero telemetry, no analytics, no phone-home. Your transcriptions, audio, and usage stats stay on your machine, stored locally in SQLite.
Where are my cloud API keys stored?
If you opt into a cloud engine (bring your own key), API keys are stored only in your operating system's keyring, never on disk or in localStorage.
Which platforms are supported?
macOS 10.15 and newer, Windows, and Linux, including first-class Wayland support with pure-Rust in-process text injection, plus X11.
Which models can I use?
Locally: Whisper GGML and GGUF from tiny to large, Parakeet TDT v3, Moonshine, Wav2Vec2 and CTC models, and more. In the cloud: OpenAI, OpenRouter, Google Gemini, or any OpenAI-compatible endpoint. (Anthropic Claude is not offered, since it does not accept audio input.) Download models in the built-in manager or import your own.
How do I install it, and what permissions are needed?
Download the installer for your OS from GitHub Releases. On first launch, download a local model or add a cloud key. On macOS, grant Microphone and Accessibility permissions; Accessibility is required for auto-paste.
What are the keyboard shortcuts?
Toggle recording with Ctrl / Cmd + Shift + Space from anywhere, and copy your last transcription with Ctrl / Cmd + Shift + C. Both are customizable in Settings.
Can I contribute?
Absolutely. Issues and pull requests are welcome on GitHub. Bug reports, feature ideas, and patches all help.