Repository navigation
Releases: arindamroynitw/loqui
Releases · arindamroynitw/loqui
Release list
Loqui v1.1 - Smart Commands & Custom Vocabulary
Loqui v1.1 - Smart Commands & Custom Vocabulary
What's New
🎤 Smart Voice Commands
Edit your dictated text on the fly with voice commands:
- Replace: "replace meeting with call", "change X to Y"
- Undo: "scratch that", "undo that"
- Formatting: "new paragraph", "new line"
- Capitalization: "all caps NASA", "capitalize john"
✨ Enhanced Cleanup
- List formatting: "First... Second... Third..." → "1. ... 2. ... 3. ..."
- Semantic filler detection: Preserves meaningful words (e.g., "I like pizza")
- False start removal: "I think we should, actually let's do Tuesday" → "Let's do Tuesday"
- Smart paragraph breaks: LLM-detected topic shifts
📚 Custom Vocabulary Builder
- Add proper nouns and technical terms that Whisper often mishears
- Corrections applied automatically during cleanup
- CSV import/export for easy sharing and backup
- Examples: "Neeraj", "Scapia", "Kubernetes"
Performance
Total latency: ~5-7 seconds (fn release → text appears)
- Whisper transcription: ~3s (on-device)
- Voice commands: ~0.2s (when used)
- Enhanced cleanup: ~0.3s
- Network variance: ~1-2s
Note: v1.1 adds ~1.5-3.5s overhead vs v1.0 due to additional LLM passes, but provides significantly better output quality.
Improvements
- Groq API timeout increased: 5s → 15s (more reliable on free tier)
- LLM prompts refined: Prevents hallucination and over-correction
- Settings reorganized: New Vocabulary tab for easy management
Installation
- Download
Loqui-v1.1.dmg - Open the DMG and drag Loqui to Applications
- Launch Loqui and grant permissions when prompted
- Configure API keys in Settings (Groq and/or OpenAI)
Requirements:
- macOS 14.0+ (Apple Silicon optimized)
- Groq API key (free tier available) and/or OpenAI API key
- ~600MB disk space for Whisper model
Known Issues
- Text insertion overwrites clipboard (by design for simplicity)
- First launch downloads 594MB Whisper model (one-time)
- Network connection required for LLM features (transcription works offline)
SHA256 Checksum
6e33c9a932a3dcc0f3f213e14efa7699290168253e47e6100122e0941d760e9b Loqui-v1.1.dmg
Support
- Report issues: https://github.com/arindamroynitw/loqui/issues
- Email: arindamroynitw@gmail.com
- Twitter/X: @crosschainyoda
Full Changelog: v1.0...v1.1