What's New
M4B Audiobook Export
Export your audiobook as a chaptered M4B file (AAC audio in MP4 container) compatible with Audiobookshelf, Apple Books, VLC, and other audiobook players.
- Smart chapter detection (default) — auto-detects chapter/part/prologue headings in the script and groups content into chapters
- Per-chunk chapters — toggle for fine-grained navigation where every line becomes a chapter
- Metadata support — fill in title, author, narrator, year, and description at export time
- Cover art embedding — upload a cover image that gets embedded in the M4B file
- Collapsible "M4B Export Options" panel on the Result tab
Note: Some Linux audiobook players (e.g. Cozy) have limited M4B support. Tested with VLC, Haruna, and Audiobookshelf.
Voice Cloning File Upload
Upload reference audio files directly from the Voices tab for voice cloning — no more manual file placement.
- Upload button next to each clone voice slot
- Clone prompt cache automatically invalidates when reference audio changes
External TTS Mode Improvements
- Render button now routes through parallel workers in external TTS mode (matching local mode behavior)
- Fix: missing
shutilimport for external TTS mode (community fix by @Robertoge65)
Bug Fixes
- Fix cancel button not stopping generation — cancel flag now properly checked between batches; stuck "generating" chunks cleaned up on startup (#16)
- Fix progressive batch rendering slowdown —
torch.compiledynamo guard accumulation reset between batches (#15) - Fix voice config not persisting when switching designed voices in clone dropdown (#17)
- Fix LLM script crash with malformed JSON — filter non-dict entries from LLM JSON arrays to prevent
AttributeError: 'str' object has no attribute 'get'(#20) - Bind server to 127.0.0.1 instead of 0.0.0.0
Documentation
- README updated with M4B export section (English and Chinese)
- Result Tab documentation expanded with all three export options
- Output Files section updated with M4B format details
Full Changelog
Contributors
Robertoge65