babanovac1980/chatterbox-tts-server repository overview

⁠Chatterbox TTS Server Docker Image

A containerized FastAPI server for the powerful Chatterbox TTS model by Resemble AI. Features a modern Web UI, voice cloning, large text processing, and GPU acceleration.

⁠Features

🗣️ High-quality text-to-speech synthesis
🎤 Voice cloning with reference audio
🌐 Modern Web UI and REST API
📚 Large text handling with intelligent chunking
🚀 GPU acceleration (NVIDIA CUDA 12.4)
🔄 OpenAI-compatible API endpoints
📦 Pre-built voices included

⁠Quick Start

Prerequisites:

• Docker & Docker Compose
• NVIDIA GPU with CUDA 12.4+ support
• NVIDIA Container Toolkit

⁠Configuration

🔧 Environment Variables:

• PORT - Host port to expose (default: 8004)
• HF_HUB_ENABLE_HF_TRANSFER - Enable faster downloads (default: 1)
• NVIDIA_VISIBLE_DEVICES - GPU devices to expose (default: all)
• NVIDIA_DRIVER_CAPABILITIES - Driver capabilities (default: compute,utility)

📁 Volume Mounts:

• ./config/config.yaml → /app/config.yaml (Application configuration)
• ./data/voices → /app/voices (Predefined voice files)
• ./data/reference_audio → /app/reference_audio (Reference audio for cloning)
• ./data/outputs → /app/outputs (Generated audio files)
• ./data/logs → /app/logs (Application logs)
• ./data/models → /app/hf_cache (Cached model files)

⁠API Usage

Generate Speech:

curl -X POST "http://localhost:8004/tts" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Hello, this is Chatterbox TTS!",
    "voice_mode": "predefined",
    "voice_id": "default"
  }' \
  --output speech.wav

Voice Cloning:

curl -X POST "http://localhost:8004/tts" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "This will sound like the reference voice!",
    "voice_mode": "clone",
    "clone_reference_filename": "my_voice.wav"
  }' \
  --output cloned_speech.wav

⁠Configuration Options

The default configuration works out of the box. To customize:

1. Edit config/config.yaml
2. Restart the container: docker restart chatterbox-tts-server

Key configuration options:

• tts_engine.device: Set to "auto", "cuda", or "cpu"
• server.port: Internal server port (keep as 8000)
• generation_defaults: Default TTS parameters

⁠GPU Requirements

• NVIDIA GPU with Compute Capability 6.0+
• CUDA 12.4+ support
• Minimum 4GB VRAM recommended
• 8GB+ VRAM for optimal performance

⁠Troubleshooting

GPU not detected:

• Ensure NVIDIA Container Toolkit is installed
• Check GPU compatibility: nvidia-smi
• Verify Docker can access GPU: docker run --gpus all nvidia/cuda:12.4.0-base-ubuntu22.04 nvidia-smi

Model download issues:

• Check internet connection
• Ensure sufficient disk space (models are ~2-3GB)
• Check HuggingFace Hub status

Performance issues:

• Increase shared memory: add shm_size: '2gb' to docker-compose
• Monitor GPU memory usage
• Consider reducing model precision in config