Audiopub v1.0.0
Purpose: Turn EPUBs into on-device audiobooks using the bundled Supertonic ONNX TTS models—no cloud required.
Setup inputs: Place onnx models and voice style JSONs under audiopub/assets/ (defaults already included: voices F1/F2/M1/M2). Pick an EPUB, output folder, and a voice in the UI.
Generation flow: EPUB is chunked to natural text blocks, synthesized via ONNX, stitched per chapter with silence/crossfade, then muxed to a single .m4b in your output directory.
Fresh temp handling: Each run now uses a unique temp folder keyed to both book and voice, so switching voices or rerunning never reuses old chunks.
Playback built-in: After conversion finishes, the latest .m4b is streamable directly in the left panel via a native audio player (seek/volume).
Scaling fix: The UI self-reloads once on first load to settle sizing if elements render off-scale.
Server controls: A “Restart Server” button is available in the UI; output files are also served via a /media/{token} endpoint for the preview player.
Defaults: Speech speed/steps are set in audiopub/config.py (DEFAULT_SPEED, DEFAULT_STEPS), with sensible chunking and silence settings for smooth listening.
Running: Activate the virtualenv and launch with python -m audiopub.main; logs go to /tmp/audiopub.log.