OmniTTS

vLLM-Omni is an inference service system that supports a variety of TTS models, including Qwen3-TTS, CosyVoice and VoxCpm2. For more models, see https://docs.vllm.ai/projects/vllm-omni/en/latest/serving/speech_api/. It provides a unified TTS API that complies with OpenAI specifications, and supports uploading, deleting and viewing cloned voices.

1. Install Dependencies

uv venv --python 3.12 --seed
source .venv/bin/activate
uv pip install vllm==0.22.0 --torch-backend=auto

2. Install vllm-omni

git clone https://github.com/vllm-project/vllm-omni.git

cd vllm-omni
uv pip install -e .

3. Start the API Service

vllm serve Qwen/Qwen3-TTS-12Hz-0.6B-Base --omni --port 8091 # Replace with the corresponding model name if using other TTS models