CosyVoice声音克隆

仓库地址：https://github.com/FunAudioLLM/CosyVoice

下载代码 git clone https://ghproxy.net/https://github.com/FunAudioLLM/CosyVoice

git clone --recursive https://github.com/FunAudioLLM/CosyVoice.git
cd CosyVoice
git submodule update --init --recursive

使用miniconda安装

conda create -n cosyvoice python=3.10
conda activate cosyvoice

此次可能安装出错，需要安装一下依赖，多执行几次，确保依赖全部安装完成。

pip install -r requirements.txt -i https://mirrors.aliyun.com/pypi/simple/ --trusted-host=mirrors.aliyun.com

cuda环境根据需求安装，这里选择cuda11.8

pip下载pytorch太慢，可先用IDM下载pytorch

https://download.pytorch.org/whl/cu118/torch-2.5.1%2Bcu118-cp311-cp310-win_amd64.whl

然后安装

pip install -U G:\pytorch路径\torch-2.5.1+cu118-cp311-cp311-win_amd64.whl torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

这两个安装多执行几次，确保安装完成

使用python执行下面程序,下载模型

from modelscope import snapshot_download
snapshot_download('iic/CosyVoice-300M', local_dir='pretrained_models/CosyVoice-300M')
snapshot_download('iic/CosyVoice-ttsfrd', local_dir='pretrained_models/CosyVoice-ttsfrd')

然后

cd pretrained_models/CosyVoice-ttsfrd/目录
解压 resource.zip
pip install ttsfrd_dependency-0.1-py3-none-any.whl
cd ../..

启动网页服务

python3 webui.py --port 50000 --model_dir pretrained_models/CosyVoice-300M

服务启动起来后，使用浏览器打开http://127.0.0.1:5000