GPT-SoVITS Microservice¶

GPT-SoVITS allows you to to do zero-shot voice cloning and text to speech of multi languages such as English, Japanese, Korean, Cantonese and Chinese.

This microservice is validated on Xeon/CUDA. HPU support is under development.

Build the Image¶

docker build -t opea/gpt-sovits:latest --build-arg http_proxy=$http_proxy --build-arg https_proxy=$https_proxy -f comps/tts/gpt-sovits/Dockerfile .

Start the Service¶

docker run  -itd -p 9880:9880 -e http_proxy=$http_proxy -e https_proxy=$https_proxy opea/gpt-sovits:latest

Test¶

Chinese only

curl localhost:9880/ -XPOST -d '{
    "text": "先帝创业未半而中道崩殂，今天下三分，益州疲弊，此诚危急存亡之秋也。",
    "text_language": "zh"
}' --output out.wav

English only

curl localhost:9880/ -XPOST -d '{
    "text": "Discuss the evolution of text-to-speech (TTS) technology from its early beginnings to the present day. Highlight the advancements in natural language processing that have contributed to more realistic and human-like speech synthesis. Also, explore the various applications of TTS in education, accessibility, and customer service, and predict future trends in this field. Write a comprehensive overview of text-to-speech (TTS) technology.",
    "text_language": "en"
}' --output out.wav

Auto detection of languages

curl localhost:9880/ -XPOST -d '{
    "text": "Hi 你好，这里是一个 cross-lingual 的例子。",
    "text_language": "auto"
}' --output out.wav

Change reference audio

curl localhost:9880/change_refer -d '{
    "refer_wav_path": "path_to_your_audio.wav",
    "prompt_text": "transcription_of_your_audio",
    "prompt_language": "language_of_your_audio"
}'