19:I[15621,["/_next/static/immutable/chunks/3gm-8e62xn1qd.js","/_next/static/immutable/chunks/0n2yq6f-f75k2.js"],"ArticleReader"] 1a:T581,

Fri Jun 20 03:45:00 2025
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 535.104.05   Driver Version: 535.104.05   CUDA Version: 12.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MEX    |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 00000000:01:00.0 Off |                  N/A |
| N/A   38C    P8    12W / 250W |      0MiB / 11264MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
| No running processes found                                                  |
+-----------------------------------------------------------------------------+

mkdir -p tools/asr/models/FasterWhisper-large-v3
wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=config.json" -O tools/asr/models/FasterWhisper-large-v3/config.json
wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=configuration.json" -O tools/asr/models/FasterWhisper-large-v3/configuration.json
wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=model.bin" -O tools/asr/models/FasterWhisper-large-v3/model.bin
wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=preprocessor_config.json" -O tools/asr/models/FasterWhisper-large-v3/preprocessor_config.json
wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=README.md" -O tools/asr/models/FasterWhisper-large-v3/README.md
wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=tokenizer.json" -O tools/asr/models/FasterWhisper-large-v3/tokenizer.json
wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=vocabulary.json" -O tools/asr/models/FasterWhisper-large-v3/vocabulary.json

1d:T566,mkdir -p tools/asr/models/FasterWhisper-large-v3 wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=config.json" -O tools/asr/models/FasterWhisper-large-v3/config.json wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=configuration.json" -O tools/asr/models/FasterWhisper-large-v3/configuration.json wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=model.bin" -O tools/asr/models/FasterWhisper-large-v3/model.bin wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=preprocessor_config.json" -O tools/asr/models/FasterWhisper-large-v3/preprocessor_config.json wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=README.md" -O tools/asr/models/FasterWhisper-large-v3/README.md wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=tokenizer.json" -O tools/asr/models/FasterWhisper-large-v3/tokenizer.json wget -c "https://www.modelscope.cn/api/v1/models/keepitsimple/faster-whisper-large-v3/repo?Revision=master&FilePath=vocabulary.json" -O tools/asr/models/FasterWhisper-large-v3/vocabulary.json11:["$","main",null,{"className":"reader-page","children":[["$","div",null,{"className":"reader-kicker","children":[["$","$L10",null,{"href":"/sources/07e5d58fbe5ad069","children":[["$","svg",null,{"ref":"$undefined","xmlns":"http://www.w3.org/2000/svg","width":16,"height":16,"viewBox":"0 0 24 24","fill":"none","stroke":"currentColor","strokeWidth":2,"strokeLinecap":"round","strokeLinejoin":"round","className":"lucide lucide-arrow-left","children":[["$","path","1l729n",{"d":"m12 19-7-7 7-7"}],["$","path","x3x0zl",{"d":"M19 12H5"}],"$undefined"]}]," 返回 ","keggin's blog"]}],["$","span",null,{"children":"工程"}]]}],["$","header",null,{"className":"article-meta","children":[["$","span",null,{"children":["FROM · ","keggin's blog"]}],["$","div",null,{"children":[["$","svg",null,{"ref":"$undefined","xmlns":"http://www.w3.org/2000/svg","width":15,"height":15,"viewBox":"0 0 24 24","fill":"none","stroke":"currentColor","strokeWidth":2,"strokeLinecap":"round","strokeLinejoin":"round","className":"lucide lucide-calendar-days","children":[["$","path","1cmpym",{"d":"M8 2v4"}],["$","path","4m81vk",{"d":"M16 2v4"}],["$","rect","1hopcy",{"width":"18","height":"18","x":"3","y":"4","rx":"2"}],["$","path","8toen8",{"d":"M3 10h18"}],["$","path","6423bh",{"d":"M8 14h.01"}],["$","path","1etili",{"d":"M12 14h.01"}],["$","path","1gbofw",{"d":"M16 14h.01"}],["$","path","lrp35t",{"d":"M8 18h.01"}],["$","path","mhygvu",{"d":"M12 18h.01"}],["$","path","kzsmim",{"d":"M16 18h.01"}],"$undefined"]}]," ","2025年6月20日"," · Keggin"]}]]}],["$","$L19",null,{"article":{"title":"GPT-SoVITS 安装指南：在Ubuntu上使用Conda配置Python环境、安装ffmpeg、CUDA检测与模型下载","byline":"Keggin","publishedAt":"2025-06-20T05:48:00.000Z","language":"fr","blocks":[{"id":"b0","type":"paragraph","html":"

GPT-SoVITS-WebUI是一个强大的少样本语音转换与语音合成 Web 用户界面.

","text":"GPT-SoVITS-WebUI是一个强大的少样本语音转换与语音合成 Web 用户界面."},{"id":"b1","type":"list","html":"

零样本文本到语音 (TTS): 输入 5 秒的声音样本, 即刻体验文本到语音转换.
少样本 TTS: 仅需 1 分钟的训练数据即可微调模型, 提升声音相似度和真实感.
跨语言支持: 支持与训练数据集不同语言的推理, 目前支持英语、日语、韩语、粤语和中文.
WebUI 工具: 集成工具包括声音伴奏分离、自动训练集分割、中文自动语音识别 (ASR) 和文本标注, 协助初学者创建训练数据集和 GPT/SoVITS 模型.

","text":"零样本文本到语音 (TTS): 输入 5 秒的声音样本, 即刻体验文本到语音转换.少样本 TTS: 仅需 1 分钟的训练数据即可微调模型, 提升声音相似度和真实感.跨语言支持: 支持与训练数据集不同语言的推理, 目前支持英语、日语、韩语、粤语和中文.WebUI 工具: 集成工具包括声音伴奏分离、自动训练集分割、中文自动语音识别 (ASR) 和文本标注, 协助初学者创建训练数据集和 GPT/SoVITS 模型."},{"id":"b2","type":"paragraph","html":"

项目原地址:RVC-Boss/GPT-SoVITS: 1 min voice data can also be used to train a good TTS model! (few shot voice cloning)

","text":"项目原地址:RVC-Boss/GPT-SoVITS: 1 min voice data can also be used to train a good TTS model! (few shot voice cloning)"},{"id":"b3","type":"paragraph","html":"

原项目地址展示的示例:

","text":"原项目地址展示的示例:"},{"id":"b4","type":"paragraph","html":"

安装 Anaconda/Miniconda

","text":"安装 Anaconda/Miniconda"},{"id":"b5","type":"paragraph","html":"

GPT-SoVITS 项目推荐使用 conda 环境进行管理。如果尚未安装 Anaconda 或 Miniconda，请先安装.

","text":"GPT-SoVITS 项目推荐使用 conda 环境进行管理。如果尚未安装 Anaconda 或 Miniconda，请先安装."},{"id":"b6","type":"paragraph","html":"

# 下载 Miniconda 安装脚本\nwget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh\n\n# 运行安装脚本\nbash Miniconda3-latest-Linux-x86_64.sh\n\n# 激活 conda 环境\nsource ~/.bashrc

","text":"# 下载 Miniconda 安装脚本 wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh # 运行安装脚本 bash Miniconda3-latest-Linux-x86_64.sh # 激活 conda 环境 source ~/.bashrc"},{"id":"b8","type":"paragraph","html":"

创建并激活 Conda 环境

","text":"创建并激活 Conda 环境"},{"id":"b9","type":"paragraph","html":"

conda create -n GPTSoVits python=3.10\nconda activate GPTSoVits

","text":"conda create -n GPTSoVits python=3.10 conda activate GPTSoVits"},{"id":"b10","type":"paragraph","html":"

安装系统依赖:

","text":"安装系统依赖:"},{"id":"b11","type":"paragraph","html":"

项目文档中提到了 ffmpeg 和 libsox-dev，这些是 Ubuntu 系统上的依赖。如果没有预先安装，需要执行下列命令:

","text":"项目文档中提到了 ffmpeg 和 libsox-dev，这些是 Ubuntu 系统上的依赖。如果没有预先安装，需要执行下列命令:"},{"id":"b12","type":"paragraph","html":"

sudo apt update\nsudo apt install ffmpeg -y\nsudo apt install libsox-dev -y

","text":"sudo apt update sudo apt install ffmpeg -y sudo apt install libsox-dev -y"},{"id":"b14","type":"paragraph","html":"

依赖的安装依据服务器性能，大约需要 2 分钟左右。

","text":"依赖的安装依据服务器性能，大约需要 2 分钟左右。"},{"id":"b15","type":"paragraph","html":"

克隆项目仓库:

","text":"克隆项目仓库:"},{"id":"b16","type":"paragraph","html":"

git clone https://github.com/RVC-Boss/GPT-SoVITS.git\ncd GPT-SoVITS

","text":"git clone https://github.com/RVC-Boss/GPT-SoVITS.git cd GPT-SoVITS"},{"id":"b17","type":"paragraph","html":"

检测服务器 CPU 性能和 CUDA 版本:

","text":"检测服务器 CPU 性能和 CUDA 版本:"},{"id":"b18","type":"paragraph","html":"

由于该项目支持使用 GPU 进行加速推理和演化，因此有必要实现确定本服务器是否具备此功能，要确认 Ubuntu 服务器是否支持 CUDA，主要检查以下两个方面：

","text":"由于该项目支持使用 GPU 进行加速推理和演化，因此有必要实现确定本服务器是否具备此功能，要确认 Ubuntu 服务器是否支持 CUDA，主要检查以下两个方面："},{"id":"b19","type":"paragraph","html":"

1.NVIDIA GPU 硬件是否存在且正常工作。

","text":"1.NVIDIA GPU 硬件是否存在且正常工作。"},{"id":"b20","type":"paragraph","html":"

2.NVIDIA 驱动程序是否已正确安装。

","text":"2.NVIDIA 驱动程序是否已正确安装。"},{"id":"b21","type":"list","html":"

检查 NVIDIA GPU 硬件

","text":"检查 NVIDIA GPU 硬件"},{"id":"b22","type":"paragraph","html":"

首先，我们需要确认服务器是否安装了 NVIDIA 显卡。您可以使用 lspci 命令来列出所有 PCI 设备，并过滤出与 NVIDIA 相关的设备。

","text":"首先，我们需要确认服务器是否安装了 NVIDIA 显卡。您可以使用 lspci 命令来列出所有 PCI 设备，并过滤出与 NVIDIA 相关的设备。"},{"id":"b23","type":"paragraph","html":"

lspci | grep -i nvidia

","text":"lspci | grep -i nvidia"},{"id":"b24","type":"paragraph","html":"

如果服务器安装了 NVIDIA 显卡，可能会看到类似以下的输出：

","text":"如果服务器安装了 NVIDIA 显卡，可能会看到类似以下的输出："},{"id":"b25","type":"paragraph","html":"

01:00.0 VGA compatible controller: NVIDIA Corporation GP102 [GeForce GTX 1080 Ti] (rev a1)\n01:00.1 Audio device: NVIDIA Corporation GP102 HDMI Audio Controller (rev a1)

","text":"01:00.0 VGA compatible controller: NVIDIA Corporation GP102 [GeForce GTX 1080 Ti] (rev a1) 01:00.1 Audio device: NVIDIA Corporation GP102 HDMI Audio Controller (rev a1)"},{"id":"b26","type":"paragraph","html":"

如果没有任何输出，或者输出中没有包含“NVIDIA”字样，那么您的服务器可能没有安装 NVIDIA 显卡，或者显卡没有被系统识别。在这种情况下，您的服务器将无法支持 CUDA。那么可以不需要进行下一步检测，直接选择使用 CPU 进行计算，缺失了加速推理和计算的过程.

","text":"如果没有任何输出，或者输出中没有包含“NVIDIA”字样，那么您的服务器可能没有安装 NVIDIA 显卡，或者显卡没有被系统识别。在这种情况下，您的服务器将无法支持 CUDA。那么可以不需要进行下一步检测，直接选择使用 CPU 进行计算，缺失了加速推理和计算的过程."},{"id":"b27","type":"paragraph","html":"

2. 检查 NVIDIA 驱动程序

","text":"2. 检查 NVIDIA 驱动程序"},{"id":"b28","type":"paragraph","html":"

如果服务器安装了 NVIDIA 显卡，下一步是检查 NVIDIA 驱动程序是否已正确安装并加载。可以使用 nvidia-smi 命令来查看 NVIDIA GPU 的状态和驱动程序信息。nvidia-smi 是 NVIDIA System Management Interface 的缩写，它是一个命令行实用程序，用于监控和管理 NVIDIA GPU 设备。

","text":"如果服务器安装了 NVIDIA 显卡，下一步是检查 NVIDIA 驱动程序是否已正确安装并加载。可以使用 nvidia-smi 命令来查看 NVIDIA GPU 的状态和驱动程序信息。nvidia-smi 是 NVIDIA System Management Interface 的缩写，它是一个命令行实用程序，用于监控和管理 NVIDIA GPU 设备。"},{"id":"b29","type":"paragraph","html":"

nvidia-smi

","text":"nvidia-smi"},{"id":"b30","type":"paragraph","html":"

如果 NVIDIA 驱动程序已正确安装，可以看到类似以下的输出，其中包含驱动版本、CUDA 版本、GPU 温度、内存使用情况等信息：

","text":"如果 NVIDIA 驱动程序已正确安装，可以看到类似以下的输出，其中包含驱动版本、CUDA 版本、GPU 温度、内存使用情况等信息："},{"id":"b31","type":"paragraph","html":"$1a","text":"$1b"},{"id":"b32","type":"paragraph","html":"

检查 CUDA Toolkit 安装情况

","text":"检查 CUDA Toolkit 安装情况"},{"id":"b33","type":"paragraph","html":"

nvidia-smi 命令的输出中会显示 CUDA Version，这表示驱动程序支持的最高 CUDA 版本。但这并不意味着 CUDA Toolkit 已经安装。CUDA Toolkit 是一个开发环境，包含了编译器、库和工具，用于在 NVIDIA GPU 上进行并行计算。

","text":"nvidia-smi 命令的输出中会显示 CUDA Version，这表示驱动程序支持的最高 CUDA 版本。但这并不意味着 CUDA Toolkit 已经安装。CUDA Toolkit 是一个开发环境，包含了编译器、库和工具，用于在 NVIDIA GPU 上进行并行计算。"},{"id":"b34","type":"paragraph","html":"

要检查 CUDA Toolkit 是否安装，以及其版本，可以尝试查找 nvcc 编译器。nvcc 是 CUDA C++ 编译器。

","text":"要检查 CUDA Toolkit 是否安装，以及其版本，可以尝试查找 nvcc 编译器。nvcc 是 CUDA C++ 编译器。"},{"id":"b35","type":"paragraph","html":"

which nvcc nvcc --version

","text":"which nvcc nvcc --version"},{"id":"b36","type":"paragraph","html":"

如果 CUDA Toolkit 已安装，可能会看到类似以下的输出：

","text":"如果 CUDA Toolkit 已安装，可能会看到类似以下的输出："},{"id":"b37","type":"paragraph","html":"

/usr/local/cuda/bin/nvcc nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2023 NVIDIA Corporation Built on Tue_Aug_15_22:02:13_PDT_2023 Cuda compilation tools, release 12.2, V12.2.140 Build cuda_12.2.r535.104-cf28d39a

","text":"/usr/local/cuda/bin/nvcc nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2023 NVIDIA Corporation Built on Tue_Aug_15_22:02:13_PDT_2023 Cuda compilation tools, release 12.2, V12.2.140 Build cuda_12.2.r535.104-cf28d39a"},{"id":"b38","type":"paragraph","html":"

如果 which nvcc 没有输出，或者 nvcc --version 报错，则表示 CUDA Toolkit 没有安装或者没有正确配置环境变量。在这种情况下，则需要根据您的 Ubuntu 版本和 NVIDIA 驱动版本安装 CUDA Toolkit。可以访问 NVIDIA 官方 CUDA Toolkit 下载页面获取详细的安装指南。

","text":"如果 which nvcc 没有输出，或者 nvcc --version 报错，则表示 CUDA Toolkit 没有安装或者没有正确配置环境变量。在这种情况下，则需要根据您的 Ubuntu 版本和 NVIDIA 驱动版本安装 CUDA Toolkit。可以访问 NVIDIA 官方 CUDA Toolkit 下载页面获取详细的安装指南。"},{"id":"b40","type":"paragraph","html":"

出现以上信息则代表该服务器不具备 GPU 推理加速，只能使用 CPU 计算。

","text":"出现以上信息则代表该服务器不具备 GPU 推理加速，只能使用 CPU 计算。"},{"id":"b41","type":"paragraph","html":"

运行安装脚本

","text":"运行安装脚本"},{"id":"b42","type":"paragraph","html":"

项目提供了一个 install.sh 脚本来简化安装过程。您需要根据您的设备类型选择 --device 参数，并选择 --source 参数。

","text":"项目提供了一个 install.sh 脚本来简化安装过程。您需要根据您的设备类型选择 --device 参数，并选择 --source 参数。"},{"id":"b43","type":"paragraph","html":"

•--device:

","text":"•--device:"},{"id":"b44","type":"paragraph","html":"

•CU126: 适用于 CUDA 12.6

","text":"•CU126: 适用于 CUDA 12.6"},{"id":"b45","type":"paragraph","html":"

•CU128: 适用于 CUDA 12.8

","text":"•CU128: 适用于 CUDA 12.8"},{"id":"b46","type":"paragraph","html":"

•ROCM: 适用于 ROCm

","text":"•ROCM: 适用于 ROCm"},{"id":"b47","type":"paragraph","html":"

•CPU: 适用于 CPU

","text":"•CPU: 适用于 CPU"},{"id":"b48","type":"paragraph","html":"

•--source:

","text":"•--source:"},{"id":"b49","type":"paragraph","html":"

•HF: 从 Hugging Face 下载模型

","text":"•HF: 从 Hugging Face 下载模型"},{"id":"b50","type":"paragraph","html":"

•HF-Mirror: 从 Hugging Face 镜像下载模型

","text":"•HF-Mirror: 从 Hugging Face 镜像下载模型"},{"id":"b51","type":"paragraph","html":"

•ModelScope: 从 ModelScope 下载模型

","text":"•ModelScope: 从 ModelScope 下载模型"},{"id":"b52","type":"paragraph","html":"

应当根据实际情况选择合适的参数。例如，如果服务器支持 CUDA 12.8，并且希望从 Hugging Face 下载模型，命令如下：

","text":"应当根据实际情况选择合适的参数。例如，如果服务器支持 CUDA 12.8，并且希望从 Hugging Face 下载模型，命令如下："},{"id":"b53","type":"paragraph","html":"

bash install.sh --device CU128 --source HF

","text":"bash install.sh --device CU128 --source HF"},{"id":"b54","type":"paragraph","html":"

如果需要下载 UVR5 模型（人声 / 伴奏分离），可以添加 --download-uvr5 参数：

","text":"如果需要下载 UVR5 模型（人声 / 伴奏分离），可以添加 --download-uvr5 参数："},{"id":"b55","type":"paragraph","html":"

bash install.sh --device CU128 --source HF --download-uvr5

","text":"bash install.sh --device CU128 --source HF --download-uvr5"},{"id":"b56","type":"paragraph","html":"

由于本服务器不具备 GPU 推理的功能，因此采用 CPU-HF 的架构部署:

","text":"由于本服务器不具备 GPU 推理的功能，因此采用 CPU-HF 的架构部署:"},{"id":"b57","type":"paragraph","html":"

注意此过程需要下载大量文件，因此需要消耗大量时间，大约 30mins(保持 20MB/s 左右的下载速度实测)

","text":"注意此过程需要下载大量文件，因此需要消耗大量时间，大约 30mins(保持 20MB/s 左右的下载速度实测)"},{"id":"b58","type":"paragraph","html":"

TIPS:

","text":"TIPS:"},{"id":"b59","type":"paragraph","html":"

本人在实测时遇到了如下报错:

","text":"本人在实测时遇到了如下报错:"},{"id":"b60","type":"paragraph","html":"

Traceback (most recent call last):\n  File \"<string>\", line 1, in <module>\n  File \"/root/miniconda3/envs/GPTSoVits/lib/python3.10/site-packages/pyopenjtalk/__init__.py\", line 26, in <module>\n    from .openjtalk import OpenJTalk\nImportError: /root/miniconda3/envs/GPTSoVits/bin/../lib/libstdc++.so.6: version `GLIBCXX_3.4.32' not found (required by /root/miniconda3/envs/GPTSoVits/lib/python3.10/site-packages/pyopenjtalk/openjtalk.cpython-310-x86_64-linux-gnu.so)\n[ERROR]: Command \"PYOPENJTALK_PREFIX=$(python -c \"import os, pyopenjtalk; print(os.path.dirname(pyopenjtalk.__file__))\")\" Failed at Line 341 with Exit Code 1\n[ERROR]: Call Stack:\n  in main() at install.sh:341

","text":"Traceback (most recent call last): File \"\", line 1, in File \"/root/miniconda3/envs/GPTSoVits/lib/python3.10/site-packages/pyopenjtalk/__init__.py\", line 26, in from .openjtalk import OpenJTalk ImportError: /root/miniconda3/envs/GPTSoVits/bin/../lib/libstdc++.so.6: version `GLIBCXX_3.4.32' not found (required by /root/miniconda3/envs/GPTSoVits/lib/python3.10/site-packages/pyopenjtalk/openjtalk.cpython-310-x86_64-linux-gnu.so) [ERROR]: Command \"PYOPENJTALK_PREFIX=$(python -c \"import os, pyopenjtalk; print(os.path.dirname(pyopenjtalk.__file__))\")\" Failed at Line 341 with Exit Code 1 [ERROR]: Call Stack: in main() at install.sh:341"},{"id":"b61","type":"paragraph","html":"

这可能是因为 Conda 环境中的库版本与系统库版本存在冲突 Conda 环境中的 pyopenjtalk 模块对 libstdc++ 的特定版本有严格要求，并且没有正确链接到系统或 Conda 中更新的库。

","text":"这可能是因为 Conda 环境中的库版本与系统库版本存在冲突 Conda 环境中的 pyopenjtalk 模块对 libstdc++ 的特定版本有严格要求，并且没有正确链接到系统或 Conda 中更新的库。"},{"id":"b62","type":"paragraph","html":"

因此我们在 Conda 环境中安装 libstdcxx-ng，它通常能提供更兼容的 C++ 标准库版本。请您运行以下命令：

","text":"因此我们在 Conda 环境中安装 libstdcxx-ng，它通常能提供更兼容的 C++ 标准库版本。请您运行以下命令："},{"id":"b63","type":"paragraph","html":"

conda install -c conda-forge libstdcxx-ng -y

","text":"conda install -c conda-forge libstdcxx-ng -y"},{"id":"b64","type":"paragraph","html":"

随后继续执行以下命令，重新安装:

","text":"随后继续执行以下命令，重新安装:"},{"id":"b65","type":"paragraph","html":"

bash install.sh --device CPU --source HF

","text":"bash install.sh --device CPU --source HF"},{"id":"b67","type":"paragraph","html":"

出现了如上的提示代表 python 和 C/C++ 的依赖全部安装成功。

","text":"出现了如上的提示代表 python 和 C/C++ 的依赖全部安装成功。"},{"id":"b68","type":"paragraph","html":"

模型的安装

","text":"模型的安装"},{"id":"b69","type":"paragraph","html":"

首先安装Git LFS，其可以用来用于处理大型文件

","text":"首先安装Git LFS，其可以用来用于处理大型文件"},{"id":"b70","type":"paragraph","html":"

sudo apt update\nsudo apt install git-lfs -y

","text":"sudo apt update sudo apt install git-lfs -y"},{"id":"b71","type":"paragraph","html":"

安装完成后，运行以下命令将 GPT-SoVITS 的预训练模型应该已经通过 Git LFS 下载到您的 GPT_SoVITS/pretrained_models 目录中：

","text":"安装完成后，运行以下命令将 GPT-SoVITS 的预训练模型应该已经通过 Git LFS 下载到您的 GPT_SoVITS/pretrained_models 目录中："},{"id":"b72","type":"paragraph","html":"

git lfs install\ngit pull

","text":"git lfs install git pull"},{"id":"b73","type":"paragraph","html":"

UVR5 Weights (人声 / 伴奏分离和混响移除，额外功能)

","text":"UVR5 Weights (人声 / 伴奏分离和混响移除，额外功能)"},{"id":"b74","type":"paragraph","html":"

mkdir -p tools/uvr5/uvr5_weights\nwget -O tools/uvr5/uvr5_weights/HP_filter.pth https://huggingface.co/RVC-Boss/GPT-SoVITS/resolve/main/tools/uvr5/uvr5_weights/HP_filter.pth\n# 您需要对该页面下的所有 .pth 文件重复此操作\n

","text":"mkdir -p tools/uvr5/uvr5_weights wget -O tools/uvr5/uvr5_weights/HP_filter.pth https://huggingface.co/RVC-Boss/GPT-SoVITS/resolve/main/tools/uvr5/uvr5_weights/HP_filter.pth # 您需要对该页面下的所有 .pth 文件重复此操作"},{"id":"b75","type":"paragraph","html":"

中文 ASR (额外功能)

","text":"中文 ASR (额外功能)"},{"id":"b76","type":"paragraph","html":"

mkdir -p tools/asr/models\nwget \"https://www.modelscope.cn/api/v1/models/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online/repo?Revision=master&FilePath=model.pt\" -O tools/asr/models/Damo_ASR_model.pt\nwget \"https://www.modelscope.cn/api/v1/models/damo/speech_fsmn_vad_zh-cn-16k-common-trie-3/repo?Revision=master&FilePath=model.pt\" -O tools/asr/models/Damo_VAD_model.pt\nwget \"https://www.modelscope.cn/api/v1/models/damo/punc_ct-transformer_cn-en-common-vocab471067-large/repo?Revision=master&FilePath=model.pt\" -O tools/asr/models/Damo_Punc_model.pt

","text":"mkdir -p tools/asr/models wget \"https://www.modelscope.cn/api/v1/models/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online/repo?Revision=master&FilePath=model.pt\" -O tools/asr/models/Damo_ASR_model.pt wget \"https://www.modelscope.cn/api/v1/models/damo/speech_fsmn_vad_zh-cn-16k-common-trie-3/repo?Revision=master&FilePath=model.pt\" -O tools/asr/models/Damo_VAD_model.pt wget \"https://www.modelscope.cn/api/v1/models/damo/punc_ct-transformer_cn-en-common-vocab471067-large/repo?Revision=master&FilePath=model.pt\" -O tools/asr/models/Damo_Punc_model.pt"},{"id":"b78","type":"paragraph","html":"

此包大约需要 1.05G, 安装完成后的提示如上图所示。

","text":"此包大约需要 1.05G, 安装完成后的提示如上图所示。"},{"id":"b79","type":"paragraph","html":"

英语日语 ASR 模型下载

","text":"英语日语 ASR 模型下载"},{"id":"b80","type":"paragraph","html":"$1c","text":"$1d"},{"id":"b81","type":"paragraph","html":"

此包大约需要 2.88G, 安装完成后的提示如下图所示。

","text":"此包大约需要 2.88G, 安装完成后的提示如下图所示。"},{"id":"b83","type":"paragraph","html":"

启动 WEBUI 程序:

","text":"启动 WEBUI 程序:"},{"id":"b84","type":"paragraph","html":"

在 ~/GPT-SoVITS/ 目录下运行以下命令：

","text":"在 ~/GPT-SoVITS/ 目录下运行以下命令："},{"id":"b85","type":"paragraph","html":"

python webui.py

","text":"python webui.py"},{"id":"b86","type":"paragraph","html":"

想直接启动推理 WebUI，可以运行：

","text":"想直接启动推理 WebUI，可以运行："},{"id":"b87","type":"paragraph","html":"

python GPT_SoVITS/inference_webui.py

","text":"python GPT_SoVITS/inference_webui.py"}],"source":"$0:f:0:1:0:props:children:props:children:1:props:children:props:children:props:children:0:props:sources:322","originalUrl":"https://keggin.tech/archives/gpt-sovits-ubuntu-conda-guide","excerpt":"GPT-SoVITS-WebUI是一个强大的少样本语音转换与语音合成Web用户界面. 零样本文本到语音 (TTS): 输入 5 秒的声音样本, 即刻体验文本到语音转换. 少样本 TTS: 仅需 1 分钟的训练数据即可微调模型, 提升声音相似度和真实感. 跨语言支持: 支持与训练数据集不同语言的推理,"},"shouldTranslate":true}],"$L1e"]}]