A more reliable generation loop
Focused improvements to long-form runs and local diagnostics.
- Safer output handling with explicit overwrite behavior
- Improved low-disk-space checks before generation
- Clearer XTTS-v2 load and chunk progress reporting
Product log
Changelog
Small releases, concrete improvements. A record of how the local generation workflow is getting better.
Voice workflows, refined
A sharper command surface for reusable local voices.
- Added voice info and remove workflows
- Expanded doctor checks for FFmpeg, CUDA, and model cache
- Improved configuration output and error guidance
The first local pipeline
The initial public shape of LLMVoice.
- Local XTTS-v2 voice cloning from reference audio
- Long transcript chunking, merging, and MP3 encoding
- Voice registry, configuration, doctor, dry-run, and debug commands