TTS Batch Pipeline

This project uses the Kokoro TTS model to convert .txt files into natural-sounding speech. It processes all input files in batch, saves .wav outputs, and organizes processed data cleanly.

Directory Structure

.
├── audio.py          # Main batch processing script
├── launch.sh         # Environment setup and runner
├── requirements.txt  # Python dependencies
├── input/            # Put your .txt files here
├── output/           # Generated .wav files appear here
└── processed/        # Processed .txt files are moved here

Usage

1. Add input files
   Place .txt files into the input/ directory.

2. Run the pipeline
   Make sure launch.sh is executable:

   chmod +x launch.sh
   ./launch.sh

   The script will:

   • Set up a Python virtual environment
   • Install dependencies
   • Detect and use GPU if available
   • Run audio.py to process all .txt files

3. Results

   • Audio is saved as .wav in output/
   • Input files are moved to processed/

Dependencies

All dependencies are listed in requirements.txt. GPU support is handled automatically in launch.sh.

Notes

• Output sample rate: 24 kHz
• Voice used: af_heart
• If no GPU is available, falls back to CPU.