Optimizing generation effects can be done in three ways:
Efficiency gains
- Use multi-threaded acceleration: e.g.
python app/createAudio.py --threads 20command to start 20 threads. - Distributed processing: measured 5 machines in parallel can process 2000 chapters in 5 hours.
Quality improvement
- Replacement of TTS engine: default CosyVoice2-0.5B has limited effect, better speech synthesis model can be integrated.
- Manual review: by
gui.pytool to check audio order and integrity.
error handling
- Re-run in case of network outage
getZjList.pyCrawl through the missing chapters. - Consider using a proxy server to switch IPs when you encounter IP restrictions.
Note: In silico models have API call limitations, and large-scale generation requires reasonable planning of task scheduling.
This answer comes from the articleTool to automatically crawl novels and generate multi-character audiobooksThe
May not be reproduced without permission:AI productivity tools " How can I improve efficiency and quality when generating audiobooks with Auto-Audio-Book?
English 
简体中文 
日本語 
Deutsch 
Português do Brasil