IndexTTS FastAPI Service

This repository provides a FastAPI implementation for serving the IndexTTS text-to-speech model through a RESTful API. It allows you to generate high-quality speech from text using a range of voice references.

Prerequisites

• Python 3.10 or higher
• CUDA-compatible GPU (recommended) or CPU
• FFmpeg installed on your system (for audio processing)

Installation

1. Clone the Original Repository (if you haven't already)

git clone https://github.com/index-tts/index-tts.git
cd index-tts

2. Install PyTorch with the Correct GPU Architecture

Before installing the IndexTTS package, you need to install PyTorch with support for your specific GPU. This is crucial for optimal performance.

For NVIDIA GPUs:

# For CUDA 12.6
pip install torch==2.6.0 torchaudio --index-url https://download.pytorch.org/whl/cu126

# For CUDA 12.4
pip install torch==2.6.0 torchaudio --index-url https://download.pytorch.org/whl/cu124

# For CUDA 11.8
pip install torch==2.6.0 torchaudio --index-url https://download.pytorch.org/whl/cu118

For AMD GPUs:

# For ROCm
pip install torch==2.6.0 torchaudio --index-url https://download.pytorch.org/whl/rocm6.2.4

For CPU only:

pip install torch==2.6.0 torchaudio --index-url https://download.pytorch.org/whl/cpu

For MacOS (with MPS):

pip install torch==2.6.0 torchaudio

3. Install the IndexTTS Package

After installing PyTorch with the correct configuration, install the IndexTTS package:

pip install -e .

4. Download Models

Download the required model files using one of the following methods:

# Using huggingface-cli
export HF_ENDPOINT="https://hf-mirror.com"  # Optional, for faster downloads in some regions
huggingface-cli download IndexTeam/Index-TTS \
  bigvgan_discriminator.pth bigvgan_generator.pth bpe.model dvae.pth gpt.pth unigram_12000.vocab \
  --local-dir checkpoints

OR

# Using wget
wget https://huggingface.co/IndexTeam/Index-TTS/resolve/main/bigvgan_discriminator.pth -P checkpoints
wget https://huggingface.co/IndexTeam/Index-TTS/resolve/main/bigvgan_generator.pth -P checkpoints
wget https://huggingface.co/IndexTeam/Index-TTS/resolve/main/bpe.model -P checkpoints
wget https://huggingface.co/IndexTeam/Index-TTS/resolve/main/dvae.pth -P checkpoints
wget https://huggingface.co/IndexTeam/Index-TTS/resolve/main/gpt.pth -P checkpoints
wget https://huggingface.co/IndexTeam/Index-TTS/resolve/main/unigram_12000.vocab -P checkpoints

5. Install FastAPI Service Dependencies

Install the additional dependencies required for the FastAPI service:

pip install -r requirements.txt

Setting Up Voice References

Before using the API, you need to set up voice reference files:

1. Create a characters directory in the project root:

   mkdir -p characters
   

2. Add WAV files containing voice samples to this directory. Each file should:

   • Be in WAV format
   • Contain a clear voice sample (5-10 seconds is usually sufficient)
   • Be named according to your preferred voice identifier (e.g., alex.wav, female1.wav)

Running the FastAPI Service

Run the FastAPI service using the provided run.py script:

python run.py

By default, the server will listen on all interfaces (0.0.0.0) on port 8000.

Advanced Run Options

You can customize the server behavior with these command-line arguments:

python run.py --host 127.0.0.1 --port 9000 --log-level debug --reload

Available options:

• --host: Host to bind to (default: 0.0.0.0)
• --port: Port to bind to (default: 8000)
• --reload: Enable auto-reload for development
• --log-level: Set logging level (default: info)
• --no-fp16: Disable FP16 precision (use for compatibility with older GPUs)
• --device: Specify device to use (cpu, cuda, cuda:0, mps)

API Usage

Generating Speech

Endpoint: POST /v1/audio/speech

Headers:

• Authorization: Bearer <your_token>
• Content-Type: application/json

Request Body:

{
  "model": "IndexTTS",
  "input": "Hello, this is a test message for IndexTTS.",
  "voice": "alex",
  "response_format": "mp3",
  "sample_rate": 24000,
  "stream": false,
  "speed": 1.0,
  "gain": 0.0
}

Parameters:

• model: Always "IndexTTS"
• input: Text to synthesize
• voice: Voice identifier (filename without extension in the characters directory)
• response_format: Output audio format (mp3, wav, or ogg)
• sample_rate: Output sample rate in Hz
• stream: Whether to stream the response
• speed: Speech speed factor (1.0 = normal)
• gain: Audio gain in dB (0.0 = normal)

Example Using Python Client

A sample client is provided in client_example.py:

python client_example.py \
  --text "Hello, this is a test message for IndexTTS." \
  --voice alex \
  --output output.mp3 \
  --format mp3 \
  --sample-rate 24000

Example Using curl

curl -X POST "http://localhost:8000/v1/audio/speech" \
  -H "Authorization: Bearer test_token" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "IndexTTS",
    "input": "Hello, this is a test message for IndexTTS.",
    "voice": "alex",
    "response_format": "mp3",
    "sample_rate": 24000,
    "stream": false,
    "speed": 1.0,
    "gain": 0.0
  }' \
  --output output.mp3

Troubleshooting

GPU Issues

• If you encounter CUDA out-of-memory errors, try using a smaller batch size or reducing model precision with the --no-fp16 flag.
• Ensure you have the correct PyTorch version installed for your CUDA version. Check with python -c "import torch; print(torch.version.cuda)".

Audio Generation Issues

• Make sure FFmpeg is installed and available in your PATH.
• Check that your reference voice files are valid WAV files of reasonable length (5-10 seconds) and quality.
• For better performance with Chinese text, ensure the text is properly formatted with spaces between characters.

API Connection Issues

• Check that the API server is running and accessible at the specified host and port.
• Verify that you're including the Authorization header with a valid token.
• If using Docker, ensure ports are properly mapped.

License

This FastAPI implementation is provided according to the license terms of the original IndexTTS project. Please refer to the license files in the original repository for more information.

Acknowledgements

This FastAPI service is built on top of the IndexTTS text-to-speech system developed by the bilibili Index Team. The original repository and research paper can be found at:

• GitHub: https://github.com/index-tts/index-tts
• Paper: IndexTTS: An Industrial-Level Controllable and Efficient Zero-Shot Text-To-Speech System