hyprwhspr

Native speech-to-text for Linux - Fast, accurate, private and hackable system-wide dictation

instant performance | Cohere / Parakeet / Whisper / Gemini / ElevenLabs / REST API | stylish visuals

Supports Arch, Debian, Ubuntu, Fedora, openSUSE and more

hyprwhspr.com

https://github.com/user-attachments/assets/4c223e85-2916-494f-b7b1-766ce1bdc991

• Built for Linux - Native AUR package for Arch, or use Debian/Ubuntu/Fedora/openSUSE
• Very fast defaults - Nearly instant and accurate performance via in-memory models
• Offline and private - Transcription never leaves your machine. Cloud APIs are fully optional.
• Latest models - Cohere Transcribe? Turbo-v3? Parakeet TDT V3? Latest and greatest
• GPU memory efficient - Limit or zero memory usage easily, more for other local models
• onnx-asr for wild CPU speeds - No GPU? Optimized for great speed on any hardware
• Translation - Translate non-English to English with a single config
• REST API or websockets - Secure, fast wires to top clouds like Gemini, ElevenLabs
• Themed visualizer - Visualizes your voice, will automatch Omarchy and Noctalia themes
• Word overrides and prompts - Custom hot keys, common words, and more
• Multi-lingual - Great performance in many languages
• Long form mode with saving - Pause, think, resume, pause: submit... Bam!
• Auto-paste anywhere - Instant paste into any active buffer, or even auto enter (optional)
• Audio ducking 🦆 - Reduces or pauses other audio on record (optional)

Why hyprwhspr? There are a lotta dictation apps. This one is built ground-up for the highest-end machines: a recent Nvidia card gets the best possible accuracy and speed, and everything else runs as well as the hardware you bring. Actively maintained, fully featured, strictly offline unless you use an API, and works on both Wayland and X11.

Quick start

Prerequisites

• Linux with systemd (Arch, Debian, Ubuntu, Fedora, openSUSE, etc.)
• Wayland or X11 session (GNOME, KDE Plasma, Sway, Hyprland, Niri, etc.).
• Clipboard/window tools:
  • wl-clipboard and wtype on Wayland
  • python-pyperclip, xclip, xdotool, and xprop on X11 (installed by the dependency script)
• Waybar or Noctalia (optional, for status bar)
• gtk4 + PyCairo (optional, for visualizer)
• NVIDIA GPU (optional, for CUDA acceleration)
• AMD/Intel GPU / APU (optional, for Vulkan acceleration)

Install

Arch Linux

On the AUR:

# Install for stable
yay -S hyprwhspr

# Or install for bleeding edge
yay -S hyprwhspr-git

Then run the interactive setup:

hyprwhspr setup

Ubuntu, Debian, Fedora, openSUSE

curl -fsSL https://hyprwhspr.com/install.sh | bash

Installs dependencies, clones to ~/.local/share/hyprwhspr/src, and walks you through setup.

# Clone the repo
git clone https://github.com/goodroot/hyprwhspr.git
cd hyprwhspr

# Install dependencies for your distro
./scripts/install-deps.sh

# Run interactive setup
./bin/hyprwhspr setup

Setup then walks you through:

1. ✅ Configure transcription backend (Cohere Transcribe, Parakeet TDT V3, Whisper, REST API, or Realtime WebSocket)
2. ✅ Download models
3. ✅ Configure themed visualizer for maximum coolness (optional)
4. ✅ Configure bar integration for your shell -- Waybar or Noctalia (optional)
5. ✅ Set up systemd user services
6. ✅ Set up permissions
7. ✅ Validate installation

First use

Ensure your microphone of choice is available in audio settings!

1. Log out and back in (for group permissions)
2. Press Super+Alt+D to start dictation - beep!
3. Speak naturally
4. Press Super+Alt+D again to stop dictation - boop!
5. Bam! Text appears in active buffer!

What you'll see while recording: on layer-shell compositors (Hyprland, Sway, niri, KDE) the animated mic OSD overlay -- on Noctalia / Omarchy it auto-matches your live shell theme; on GNOME/Mutter you may need to make additional changes. See Themed visualizer for details.

Any snags, please create an issue.

Updating

# Arch: update via your AUR helper
yay -Syu hyprwhspr

# Other distros: re-run the installer
curl -fsSL https://hyprwhspr.com/install.sh | bash

# Either way, setup is idempotent if you need to re-run it
hyprwhspr setup

CLI commands

• hyprwhspr setup - Interactive initial setup
• hyprwhspr config - Manage configuration (show / show --all / edit / secondary-shortcut)
• hyprwhspr model - Manage models (download / list / status / unload / reload)
• hyprwhspr record - External hotkey control (start / stop / toggle / cancel / capture / status / copy-last / paste-last / clear-last)
• hyprwhspr transcribe INPUT - Transcribe a WAV or MP3 file to stdout (use -o PATH to write a file)
• hyprwhspr status - Overall status check
• hyprwhspr validate - Validate installation
• hyprwhspr test - Test microphone and transcription end-to-end
• hyprwhspr keyboard - List/test keyboard devices (list / test)
• hyprwhspr waybar / hyprwhspr noctalia - Manage bar integration
• hyprwhspr mic-osd - Enable/disable the mic OSD (enable / disable / status)
• hyprwhspr systemd - Manage systemd services
• hyprwhspr uninstall - Remove hyprwhspr and user data

Documentation

For full configuration and customization, see the Configuration guide.

• Minimal configuration
• Recording modes -- toggle, push-to-talk, auto, long-form
• Custom hotkeys -- key support, secondary shortcuts, Hyprland bindings
• Backends -- Cohere Transcribe, Parakeet, Whisper, REST API, Realtime WebSocket
• GPU resource management -- unload/reload model to free VRAM
• Audio and visual feedback -- visualizer, audio feedback, ducking
• Text processing -- word overrides, filler words, symbol replacements
• Paste and clipboard behavior -- paste mode, per-app paste keys, non-QWERTY, auto-submit
• Integrations -- Waybar, Noctalia, Hyprland bindings, external hotkey systems
• Troubleshooting

Getting help

• Check logs: journalctl --user -u hyprwhspr.service
• Troubleshooting covers common issues
• Still stuck? Create an issue - logs helpful!

License

MIT License - see LICENSE file.

Contributing

Start with an issue - happy to help! Pull requests best begin there too.

Upvotes on the AUR and/or GitHub stars are also a nice way to show appreciation.

Built with ❤️ in 🇨🇦