Skip to content

Green2Grey2/whipr

BranchesTags

Repository files navigation

Whipr

Whipr

Fast, private voice transcription for your desktop

Features • Installation • Usage • Configuration • Development • License

Platform License Rust

Whipr is a desktop application for fast, local voice transcription. Press a hotkey, speak, and your words are transcribed and pasted directly into any application. All processing happens on your machine—no cloud, no subscriptions, no data leaves your device.

Built with Tauri, Svelte, and whisper.cpp.

Features

  • Global Hotkeys — Start/stop recording from anywhere with customizable keyboard shortcuts
  • Local Transcription — Powered by whisper.cpp with support for multiple model sizes
  • Auto-Paste — Transcribed text is automatically inserted into your focused application
  • Transcript History — Search and browse past transcriptions with full-text search
  • GPU Acceleration — Optional CUDA, Metal, Vulkan, ROCm, and Intel oneAPI support
  • Privacy First — Everything runs locally; no internet connection required
  • Cross-Platform — Native builds for Linux, macOS, and Windows

Installation

Linux

AppImage (Recommended)

Download the latest .AppImage from Releases, then:

chmod +x Whipr_*.AppImage
./Whipr_*.AppImage

Build from Source

Prerequisites:

  • Rust (stable)
  • Node.js 18+ and npm
  • Build essentials: build-essential, pkg-config, libssl-dev
  • Audio libraries: libasound2-dev
  • WebKit: libwebkit2gtk-4.1-dev, libgtk-3-dev
  • Additional: libayatana-appindicator3-dev, librsvg2-dev

Ubuntu/Debian:

sudo apt update
sudo apt install -y build-essential pkg-config libssl-dev libasound2-dev \
  libwebkit2gtk-4.1-dev libgtk-3-dev libayatana-appindicator3-dev librsvg2-dev

Fedora:

sudo dnf install -y gcc pkg-config openssl-devel alsa-lib-devel \
  webkit2gtk4.1-devel gtk3-devel libappindicator-gtk3-devel librsvg2-devel

Arch Linux:

sudo pacman -S --needed base-devel pkg-config openssl alsa-lib \
  webkit2gtk-4.1 gtk3 libappindicator-gtk3 librsvg

Build:

git clone https://github.com/green2grey/whipr.git
cd whipr
npm install
npm run tauri build

The AppImage will be in src-tauri/target/release/bundle/appimage/.

Linux with CUDA

For NVIDIA GPU acceleration:

# Ensure CUDA toolkit is installed
npm install
npm run tauri:build:cuda

Set CUDA_ARCH to override auto-detection: CUDA_ARCH=89 npm run tauri:build:cuda

Wayland Notes

On Wayland, install clipboard and input helpers for auto-paste:

# For wtype (recommended)
sudo apt install wl-clipboard wtype

# Or ydotool
sudo apt install wl-clipboard ydotool

Global hotkeys are limited on Wayland. Use GNOME custom shortcuts or the in-app buttons instead. See GNOME Extension for an overlay solution.

macOS

DMG (Apple Silicon)

Download the latest .dmg from Releases and drag Whipr to Applications.

Build from Source

Prerequisites:

  • Xcode Command Line Tools: xcode-select --install
  • Rust (stable): curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
  • Node.js 18+: brew install node

Build:

git clone https://github.com/green2grey/whipr.git
cd whipr
npm install
npm run tauri:build:mac

The .app and .dmg will be in src-tauri/target/release/bundle/.

Note: macOS builds use Metal for GPU acceleration automatically on Apple Silicon.

Windows

Installer

Download the latest .msi or .exe installer from Releases.

Build from Source

Prerequisites:

Build:

git clone https://github.com/green2grey/whipr.git
cd whipr
npm install
npm run tauri:build:win

Installers will be in src-tauri\target\release\bundle\.

Usage

Quick Start

  1. Launch Whipr — The app starts minimized to the system tray
  2. Download a Model — Open Settings and download a transcription model (small.en recommended)
  3. Configure Hotkeys — Set your preferred keyboard shortcuts
  4. Start Recording — Press your hotkey or click the record button
  5. Speak — Your audio is captured locally
  6. Stop Recording — Press the hotkey again; text is transcribed and pasted

Hotkeys

Action Default Description
Toggle Recording Ctrl+Alt+Space Start/stop recording
Paste Last Ctrl+Alt+V Paste the last transcript
Show App Ctrl+Alt+O Bring the app window to focus

On macOS, use Cmd instead of Ctrl.

CLI Commands

Whipr supports CLI commands for integration with system shortcuts:

whipr --toggle       # Start/stop recording
whipr --paste-last   # Paste the last transcript
whipr --show         # Show the app window

GNOME Extension

For GNOME on Wayland, an optional overlay extension provides a recording indicator and quick controls:

./scripts/install-gnome-extension.sh
gnome-extensions enable whispr-overlay@greenuni

See GNOME_EXTENSION.md for details.

Configuration

Models

Whipr uses whisper.cpp models for transcription. Available models:

Model Size Speed Accuracy Use Case
tiny.en ~75 MB Fastest Good Quick notes, low-end hardware
small.en ~460 MB Fast Better Recommended default
medium.en ~1.5 GB Moderate Best High accuracy requirements

Models are downloaded to:

  • Linux: ~/.local/share/whispr/models
  • macOS: ~/Library/Application Support/whispr/models
  • Windows: %APPDATA%\whispr\models

Settings

All settings are accessible from the Settings tab:

Category Options
Audio Input device, sample rate, gain, noise gate, VAD
Hotkeys Customize all keyboard shortcuts
Transcription Model selection, GPU acceleration, custom vocabulary
Automation Auto-paste, paste delay, clipboard behavior
Storage Data location, audio retention, history cleanup

GPU Acceleration

Enable GPU acceleration in Settings for faster transcription:

Platform Backend Build Flag
NVIDIA CUDA --features cuda
AMD ROCm/HIP --features hipblas
Intel oneAPI --features intel-sycl
Apple Silicon Metal --features metal (default on macOS)
Cross-platform Vulkan --features vulkan

Development

Prerequisites

Setup

git clone https://github.com/green2grey/whipr.git
cd whipr
npm install

Development Server

# Frontend only (hot reload)
npm run dev

# Full Tauri app (recommended)
npm run tauri dev

Project Structure

whipr/
├── src/                    # Svelte frontend
│   ├── lib/               # Components and utilities
│   └── App.svelte         # Main application
├── src-tauri/             # Rust backend
│   ├── src/
│   │   ├── core/          # Audio, transcription, storage
│   │   ├── commands.rs    # Tauri IPC commands
│   │   └── main.rs        # Application entry
│   └── Cargo.toml
├── gnome-extension/       # GNOME Shell overlay
└── scripts/               # Build and install scripts

Building

# Linux (AppImage)
npm run tauri build

# Linux (CUDA)
npm run tauri:build:cuda

# macOS (DMG)
npm run tauri:build:mac

# Windows (MSI/NSIS)
npm run tauri:build:win

Troubleshooting

Common Issues

No audio input detected

  • Check that your microphone is not muted
  • Verify the correct input device is selected in Settings
  • On Linux, ensure PipeWire/PulseAudio is running

Transcription is slow

  • Try a smaller model (tiny.en)
  • Enable GPU acceleration if available
  • Check CPU usage; close resource-heavy applications

Auto-paste not working

  • On Wayland: Install wl-clipboard and wtype
  • On X11: Ensure the target application accepts Ctrl+V
  • Increase the paste delay in Settings

Model download fails

  • Check your internet connection
  • Verify write permissions to the models directory
  • Try downloading manually and placing in the models folder

License

This project is licensed under the MIT License. See LICENSE for details.

Made with Rust, Svelte, and whisper.cpp

About

Fast, private voice transcription for your desktop

Topics

transcription

Resources

Readme

License

MIT license

Contributing

Contributing
Activity
Custom properties

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases 2

v0.2.0 Latest
Feb 8, 2026
+ 1 release

Packages

 
 
 

Contributors

Languages