legendx
(ss)
1
生成的音频效果链接
腾讯云 GPU地址
https://ide.cloud.tencent.com/dashboard/gpu-workspace
先新建一个 space
流程参考
conda环境
conda create -n fish-speech python=3.10 -y
conda activate fish-speech
torch 依赖安装
pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu121
libsamplerate 和 fish-speech 依赖
前面这两步骤和csdn 文章里面是一样的, 直接使用我本地已经编译好的文件压缩包 上传到并解压
解压命令如下
tar -zxvf fish-speech.tar.gz
tar -xf libsamplerate-0.2.2.tar.xz
压缩包地址
fish speech压缩包
libsamplerate压缩包
链接:https://pan.quark.cn/s/fd2d286c0406
运行webui
命令路径 /workspace/fish-speech下
cd ./fish-speech
python -m tools.webui --llama-checkpoint-path "checkpoints/fish-speech-1.4" --decoder-checkpoint-path "checkpoints/fish-speech-1.4/firefly-gan-vq-fsq-8x1024-21hz-generator.pth" --decoder-config-name firefly_gan_vq
可以看到 http的监听端口为7860
ngork代理
打开 https://dashboard.ngrok.com/get-started/setup/linux
按 Download标签页的流程走
注意最后代理的端口是
7860 命令如下
ngrok http http://localhost:7860
最后一步
打开这个*.app链接就能使用啦
272 个赞
_lang
(lang-bit)
12
佬给的drive没有开放下载权限,可以直接参考下面的bash自行编译
# libsamp下载编译
cd /tmp
wget https://github.com/libsndfile/libsamplerate/releases/download/0.2.2/libsamplerate-0.2.2.tar.xz
tar -xvf libsamplerate-0.2.2.tar.xz
cd libsamplerate-0.2.2/
mkdir build
cd build
cmake ..
make
# fish-speech下载安装
cd /workspace
wget https://github.com/fishaudio/fish-speech/archive/refs/heads/main.zip
unzip main.zip
mv fish-speech-main fish-speech
cd ./fish-speech
apt install portaudio19-dev
pip install -e .
# 抱抱脸安装、设置镜像
pip install -U huggingface_hub
export HF_ENDPOINT=https://hf-mirror.com
# 模型下载
python ./tools/download_models.py
# 启动
python -m tools.webui --llama-checkpoint-path "checkpoints/fish-speech-1.4" --decoder-checkpoint-path "checkpoints/fish-speech-1.4/firefly-gan-vq-fsq-8x1024-21hz-generator.pth" --decoder-config-name firefly_gan_vq
下面就是自己弄一下内网穿透
4 个赞
谢谢老板,cd /workspace 脑速比手速快啊
3 个赞
legendx
(ss)
15
2 个赞
牛逼啊,这声音真像,我用nps内网穿透到自己服务器上了
3 个赞
集成到我小程序里了,“快捷配音助手”,大家可以来体验一下
4 个赞