v0.1.0-preview

VoiceInput v0.1.0-preview

⚠️ This is a PREVIEW release for testing and feedback. Expect bugs and breaking changes before stable.

Features

• Push-to-talk voice input (Alt+V)
• Language switching (Alt+L): Auto / Chinese / English
• 100% offline ASR on local GPU (Qwen3-ASR-0.6B)
• Custom terms dictionary for ASR correction
• Settings panel: microphone, hotkeys, audio, terms
• Backend crash auto-restart + file logging
• System tray with view-logs menu

System Requirements

• Windows 10 1903+ / Windows 11
• NVIDIA GPU (CUDA 11.8+), 4GB+ VRAM
• 200MB app + 1.2GB model

Install

1. Download and extract the zip
2. Run voiceinput.exe
3. First launch downloads the ASR model (~1.2GB, one-time)
4. Press Alt+V to speak, release to paste

Build from Source

See README

Known Limitations

• NSIS installer not available (sidecar is 2.7GB, exceeds NSIS mmap limit); zip distribution only
• Preview quality — please report issues