audiocoqui

A multilingual tool to convert PDF ebooks to audiobooks using XTTS v2 TTS model by cloning a speaker voice.

Features

• Clone speaker voice
• Keep sentences together
• Clean text input (remove line breaks, etc.)
• Split PDF into pages
• Convert pages to audio
• Add silence between sections
• Concatenate audio files automatically
• Progress tracking and journaling with crash recovery

Requirements

• Python 3.10+
• pip install -r requirements.txt

Setup

• Copy .env.example to .env and fill in the missing values
• Download a speaker voice sample and put it in the source_audio folder (any .wav file of more than 10 seconds should work)
• Put your PDF in the proper folder as specified in the .env file

Usage

NOTE: You can and should clean the output by removing the audio_pages folder after you're done (example in clean_output file)

• python src/main.py

Expected output

• A folder with all the audio pages of the PDF and their chunks if splitted.
• A final audiobook file as specified in the .env file.

FAQ

GPU or CPU?

While the model is capable of running on CPU, it's recommended to use a GPU for faster processing.

Model Size

The model size is slightly smaller than 2GB, so it's recommended to use a GPU with at least 4GB of VRAM or to ensure that your RAM is large enough to handle the model.

Why we use a lot of small .wav files

We use a lot of small .wav files to enable crash recovery, avoid corruption and to enable progress tracking in a more reliable way.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Credits

• Coqui XTTS v2
• PyTorch
• All the authors of the used libraries