Hold a key. Speak. Your message is sent.
Voice-to-action for Windows. No typing, no opening apps — just a hotkey and your voice.
- Download the latest release: Releases page
- Unzip anywhere (e.g.
C:\Voicy\) - Run
voicy.exe— that's it. Telegram credentials are pre-embedded. - On the Telegram tab → click QR → scan with your phone (Telegram → Settings → Devices → Link Desktop).
- Hold
Alt+Xand say "write<contact>hi" to test.
Privacy-conscious? Want to use your own Telegram app instead of ours? See TELEGRAM_SETUP.md.
Full step-by-step: INSTALL.md
After setup, press and hold Alt+X and say:
"Write Josh — running late, will be there in 10."
Release. Done. Message in your friend's Telegram inbox.
| Say | What happens |
|---|---|
write <name> <message> |
Send Telegram message |
open YouTube <query> |
Open YouTube search |
play first video |
Open first search result |
volume up 20 percent |
YouTube volume +20% |
fullscreen |
Toggle fullscreen |
pause / play |
Play/pause video |
give answer <question> |
AI assistant replies aloud |
Full list: USAGE.md
git clone git@github.com:ArtemKlunduk/Voicy.git
cd Voicy
scripts\build-release.cmdRequires Rust 1.75+ with MSVC target and Visual Studio Build Tools. Details: INSTALL.md → Building from source.
- Language: Rust 1.95 with MSVC target
- ASR: Parakeet V3 (NeMo, ONNX int8) or whisper.cpp
- MTProto: grammers-client
- AI: Google Gemini API + local LLM via candelabra (Qwen / Llama / Gemma)
- UI: wry (WebView2) + native Win32 overlay (per-pixel alpha via
UpdateLayeredWindow)
voicy/
├── src/ Rust source
├── assets/ Bundled DLLs + icon
├── docs/ User documentation
├── scripts/ Build & release scripts
├── Cargo.toml Dependencies
└── README.md
PRs welcome. See CONTRIBUTING.md.
This project is being built by two people:
- @ArtemKlunduk — Telegram, UI, AI integration
- @tuwulalo — ASR, native overlay, build
MIT — see LICENSE.