将本项目配置项修改如下即可
TTS_TYPE = "EDGE_TTS"
EDGE_TTS_VOICE = "zh-CN-XiaoxiaoNeural"
使用以下docker-compose.yml文件,先启动SoVITS服务API
- 创建以下文件夹结构
. ├── docker-compose.yml └── reference ├── clone_target_txt.txt └── clone_target_wave.mp3
- 其中
docker-compose.yml
为version: '3.8' services: gpt-sovits: image: fuqingxu/sovits_gptac_trim:latest container_name: sovits_gptac_container working_dir: /workspace/gpt_sovits_demo environment: - is_half=False - is_share=False volumes: - ./reference:/reference ports: - "19880:9880" # 19880 为 sovits api 的暴露端口,记住它 shm_size: 16G deploy: resources: reservations: devices: - driver: nvidia count: "all" capabilities: [gpu] command: bash -c "python3 api.py"
- 其中
clone_target_wave.mp3
为需要克隆的角色音频,clone_target_txt.txt
为该音频对应的文字文本( https://wiki.biligame.com/ys/%E8%A7%92%E8%89%B2%E8%AF%AD%E9%9F%B3 ) - 运行
docker-compose up
- 将本项目配置项修改如下即可
(19880 为 sovits api 的暴露端口,与docker-compose.yml中的端口对应)
TTS_TYPE = "LOCAL_SOVITS_API" GPT_SOVITS_URL = "http://127.0.0.1:19880"
- 启动本项目