【保姆教学】手把手教你本地安装，超强AI语音合成克隆工具CosyVoice，解决报错torch==2.0.1

克隆项目

git clone --recursive https://github.com/FunAudioLLM/CosyVoice.git
# If you failed to clone submodule due to network failures, please run following command until success
cd CosyVoice
git submodule update --init --recursive

如果没有conda请安装: https://docs.conda.io/en/latest/miniconda.html

创建Conda

conda create -n cosyvoice python=3.8
conda activate cosyvoice
# pynini is required by WeTextProcessing, use conda to install it as it can be executed on all platform.
conda install -y -c conda-forge pynini==2.1.5

执行requirements.txt前需要修改一下，windows系统执行会有问题，修改内容如下：

onnxruntime-gpu1.16.0; sys_platform == ‘linux’
onnxruntime1.16.0; sys_platform == ‘darwin’ or sys_platform == ‘windows’
改成
onnxruntime==1.16.0

在执行

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

执行requirements.txt，下载太慢或者报错torch==2.0.1请看这，按ctrl+c中断下载

太慢可以复制下载链接到浏览器下载，附上IDM下载器永久版：链接：https://pan.quark.cn/s/ac0eda6aaf52 提取码：uZB4

下载完成后执行安装
#安装torch-2.0.1+cu118
# H:\Downloads\torch-2.0.1+cu118-cp38-cp38-win_amd64.whl就改成自己的文件地址
pip install H:\Downloads\torch-2.0.1+cu118-cp38-cp38-win_amd64.whl -i https://pypi.tuna.tsinghua.edu.cn/simple

再次执行requirements.txt

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

执行完后
在CosyVoice目录下新建文档models.py，复制模型下载代码

# SDK模型下载
from modelscope import snapshot_download
snapshot_download('iic/CosyVoice-300M', local_dir='pretrained_models/CosyVoice-300M')
snapshot_download('iic/CosyVoice-300M-25Hz', local_dir='pretrained_models/CosyVoice-300M-25Hz')
snapshot_download('iic/CosyVoice-300M-SFT', local_dir='pretrained_models/CosyVoice-300M-SFT')
snapshot_download('iic/CosyVoice-300M-Instruct', local_dir='pretrained_models/CosyVoice-300M-Instruct')
snapshot_download('iic/CosyVoice-ttsfrd', local_dir='pretrained_models/CosyVoice-ttsfrd')

执行下载模型

 Python models.py

如果期间报错modelscope请升级，后在执行模型下载

#升级modelscope
pip install --upgrade modelscope -i https:pypi.tuna.tsinghua.edu.cn simple

最后执行运行代码，新建一个.bat文件内容可以参考下面：

@echo off
set PORT=50001
call conda activate cosyvoice
start /B python webui.py --port %PORT% --model_dir pretrained_models/CosyVoice-300M

:check_port
timeout /t 1 > nul
powershell -Command "try { (New-Object System.Net.Sockets.TcpClient).Connect('localhost', %PORT%); $true } catch { $false }" | findstr "True" > nul && (
    start http://localhost:%PORT%
    goto end
) || goto check_port

:end
pause

下载的模型有四个，分别对应：

3s极速复刻_跨语种复刻 ：使用 CosyVoice-300M 模型。
3s极速复刻_跨语种复刻_轻量 ：使用 CosyVoice-300M-25Hz 模型。
预训练音色 ：使用 CosyVoice-300M-SFT 模型。
自然语言控制 ：使用 CosyVoice-300M-Instruct 模型。

新建4个.bat文件需要修改PORT=50001端口，和启动模型pretrained_models/XXXX**

在这里插入图片描述

如果代码启动报错：
CondaError: Run ‘conda init’ before ‘conda activate’

#在命令提示符（CMD）或 PowerShell 中执行以下命令来初始化 conda：
conda init