# voice-cloning

**Repository Path**: hmsay/voice-cloning

## Basic Information

- **Project Name**: voice-cloning
- **Description**: voxcpm小白本地使用，项目https://github.com/OpenBMB/VoxCPM/
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-10-14
- **Last Updated**: 2025-10-15

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# 声音克隆

## 1. 安装 uv

```bash
winget install --id=astral-sh.uv  -e  # windows
brew install uv  # macos
curl -LsSf https://astral.sh/uv/install.sh | sh # linux or macos
```

如果不方便可以手动安装python 

[https://mirrors.aliyun.com/python-release/](https://mirrors.aliyun.com/python-release/)

## 2. 安装 git

```bash
winget install --id=Git.Git -e # windows
brew install git  # macos
sudo apt install git  # ubuntu
```

如果不方便可以手工安装git

[https://mirrors.tuna.tsinghua.edu.cn/github-release/git-for-windows/git/LatestRelease/](https://mirrors.tuna.tsinghua.edu.cn/github-release/git-for-windows/git/LatestRelease/)

## 3. 下载项目

```bash
git clone https://gitee.com/hmsay/voice-cloning.git
```

## 4. nvidia 和 cuda

具体是否可以使用gpu加速，取决于显卡是否支持，是否安装了nvidia驱动和cuda toolkit。

参考文献[2025年 Windows10 安装 CUDA Toolkit 和 cuDNN，PyTorch 安装 CUDA Runtime 和 cuDNN](https://zhuanlan.zhihu.com/p/1944361508134954370)

## 5. 安装依赖

```bash
cd voice-cloning
uv sync
```

原始的torch版本存在问题，如果需要使用gpu加速，需要安装```torch2.8.0+cu129```

```bash
卸载旧版本torch
uv pip uninstall torch
uv pip install torch==2.8.0+cu129 torchaudio -i https://mirrors.sjtug.sjtu.edu.cn/pytorch-wheels/cu129 --index-strategy unsafe-best-match
```

> 版本号需要根据显卡实际调整，例如cuda12.9对应的torch版本为2.8.0+cu129

验证torch是否安装成功

```bash
uv run python -c "import torch; print(torch.cuda.is_available())"
# True
```

## 6. 下载模型

```bash
uv run modelscope download --model iic/SenseVoiceSmall --local_dir ./models/iic/SenseVoiceSmall
uv run modelscope download --model iic/speech_zipenhancer_ans_multiloss_16k_base --local_dir ./models/iic/speech_zipenhancer_ans_multiloss_16k_base
uv run modelscope download --model openbmb/VoxCPM-0.5B --local_dir ./models/openbmb/VoxCPM-0.5B
```

## 6. 运行项目

```bash
uv run main.py
```