chichi-speech 0.1.0

Qwen3 TTS Service with Voice Cloning

Chichi Speech

A high-quality, voice-cloning TTS service powered by Qwen3.

Chichi Speech provides a robust REST API and CLI tools for text-to-speech synthesis, featuring efficient voice cloning capabilities. It is designed to be easily deployed or integrated into other AI agents and workflows.

Features

• High Quality: Utilizes the Qwen3-TTS model for state-of-the-art speech synthesis.
• Voice Cloning: Clone voices from reference audio files.
• Efficient: Optimized for reusing voice prompts to minimize computation for repeated requests.
• Standardized API: Simple REST API (/synthesize) for easy integration.
• CLI Tools: Includes chichi-speech-server and chichi-speech-client for immediate use.

Installation

Prerequisites: git, uv, python >= 3.10.

export CHICHI_SPEECH_HOME="~/chichi-speech/"
export CHICHI_SPEECH_ENV="~/chichi-speech/.venv"
git clone https://github.com/yourusername/chichi-speech.git $CHICHI_SPEECH_HOME
cd $CHICHI_SPEECH_HOME

uv venv $CHICHI_SPEECH_ENV --python 3.10
source $CHICHI_SPEECH_ENV/bin/activate

uv pip install -e .

Usage

1. Start the Service

The service runs on port 9090 by default.

# Start the server (runs in foreground, use & for background or a separate terminal)
source $$CHICHI_SPEECH_ENV/bin/activate
chichi-speech-server
# OR specify the port explicitly
chichi-speech-server --port 9090 --host 0.0.0.0
# OR specify your reference audio and text for voice cloning (Recommended)
chichi-speech-server --ref-audio /path/to/my/voice.wav --ref-text "caption of the reference audio"

2. Verify Service is Running

Check the health/docs:

curl http://localhost:9090/docs

3. Generate Speech

Use cURL:

curl -X POST "http://localhost:9090/synthesize" \
     -H "Content-Type: application/json" \
     -d '{
           "text": "Nice to meet you",
           "language": "English"
         }' \
     --output output/nice_to_meet.wav

Functionality

• Endpoint: POST /synthesize
• Default Port: 9090
• Voice Cloning: Uses a pre-computed voice prompt from reference files to ensure the cloned voice is consistent and generation is fast.

Development

Install dev dependencies:

uv pip install -e ".[dev]"

Run tests:

pytest

License

MIT