I built a simple voice-to-text app for Linux because existing solutions either don't work well or send your audio to the cloud.
What it does:
- Press hotkey → pill popup appears
- Speak your text
- Stop speaking → Ctrl+V anywhere to paste
Why it's different:
- 100% local using Whisper models
- No cloud services, no API costs
- Works system-wide (terminal, browser, IDE, everywhere)
- You choose the model size (fast vs accurate)
Built it for myself on Fedora KDE, but should work on most distros.
Check it out: https://github.com/harshsharma2455/Toice
Open to feedback and contributions!
Top comments (0)