Releases · jianchang512/pyvideotrans

v3.91更新日志

1. 修复配音渠道选择错误
2. 修复批量语音转字幕第2个以后报错问题
3. 增加 Fun-ASR-Nano-2512 语音识别模型支持
4. 增加 Huggingface_ASR 语音识别渠道，支持 moonshine 的8个模型和nvidia的一个英文模型
5. 增加 CosyVoice3 配音支持
6. 优化模型下载错误提示

v3.91 完整包下载(3.9G 含large-v3-turbo模型)

如果未安装过旧版本，请在此下载完整版，如需cuda加速，需有英伟达显卡并且安装cuda12.8及cudnn9

百度网盘下载地址: https://pan.baidu.com/s/1qk-F_sqdlF23fMTiJVrZwA?pwd=1234

huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-pyvideotrans-v3.91.7z?download=true

v3.91 补丁包(320MB)

不可单独使用，仅用于覆盖旧包，与完整包区别是不含有torch模块
因torch从 2.5 升级到 2.7，补丁包不含有torch，直接下载覆盖补丁包可能导致无法cuda加速，建议下载完整包覆盖
覆盖后如果闪退，请下载完整包

百度网盘下载地址: https://pan.baidu.com/s/1buZGYYH4Rxfc8A_OTFdbnQ?pwd=1234

huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-PatchUpdate-v3.91.7z?download=true

v3.86 Full Package Download (3.8G Includes large-v3-turbo model)

If this is a fresh installation, please download the full package. For CUDA acceleration, an NVIDIA graphics card is required, along with CUDA 12.8 and cuDNN 9.

huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-pyvideotrans-v3.86.7z?download=true

Baidu Network Disk: https://pan.baidu.com/s/1LU0y_KwNPHlfXYaELTNaTw?pwd=1234

v3.86 Patch Package (360MB)

Cannot be used standalone; this is only for overwriting a previous installation. The main difference from the full package is that it does not contain the torch module.
Since torch has been upgraded from 2.5 to 2.7, this patch does not include it. Applying the patch directly may cause CUDA acceleration to fail. It is recommended to download the full package to update.

huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-PatchUpdate-v3.86.7z?download=true

Baidu Network Disk: https://pan.baidu.com/s/1SLEDF4fqJ4tofKoU4Sc4ng?pwd=1234

Extraction Notes:

Extract the 7z archive. After extraction, find the sp.exe file and double-click to start the software. Do not run it directly from within the archive.
Do not extract the software to the desktop or folders requiring administrator permissions like Program Files on the C drive.
Avoid path errors: Do not include Chinese characters, spaces, or special symbols in the extraction path.

The current latest version is v3.85, and the patch package is v3.85. Change logs

v3.85 Full Package Download (3.8G Includes large-v3-turbo model)

If this is a fresh installation, please download the full package. For CUDA acceleration, an NVIDIA graphics card is required, along with CUDA 12.8 and cuDNN 9.

huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-pyvideotrans-v3.85.7z?download=true

Baidu Network Disk: https://pan.baidu.com/s/12r5-lQVbz-rBI0miBLosyA?pwd=1234

v3.85 Patch Package (360MB)

Cannot be used standalone; this is only for overwriting a previous installation. The main difference from the full package is that it does not contain the torch module.
Since torch has been upgraded from 2.5 to 2.7, this patch does not include it. Applying the patch directly may cause CUDA acceleration to fail. It is recommended to download the full package to update.

huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-PatchUpdate-v3.85.7z?download=true

Baidu Network Disk: https://pan.baidu.com/s/1b1fzJgDga1Zn_Qc1BKUlTg?pwd=1234

Extraction Notes:

Extract the 7z archive. After extraction, find the sp.exe file and double-click to start the software. Do not run it directly from within the archive.
Do not extract the software to the desktop or folders requiring administrator permissions like Program Files on the C drive.
Avoid path errors: Do not include Chinese characters, spaces, or special symbols in the extraction path.

Source Code Deployment

Deploying This Software via Source Code

The pre-packaged version is only for Windows 10/11. For macOS and Linux, please deploy from source
To upgrade a source deployment: pull the source code again to overwrite, then execute pip3 install --upgrade torch openai google-genai whisper elevenlabs gradio_client faster-whisper

v3.84 Full Package Download (3.8G Includes large-v3-turbo model)

If this is a fresh installation, please download the full package. For CUDA acceleration, an NVIDIA graphics card is required, along with CUDA 12.8 and cuDNN 9.

huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-videotrans-v3.83.7z?download=true

Baidu Network Disk: https://pan.baidu.com/s/1_Oz9DbpwYR9-eBmalCj7UQ?pwd=cwdm

v3.84 Patch Package (360MB)

Cannot be used standalone; this is only for overwriting a previous installation. The main difference from the full package is that it does not contain the torch module.
Since torch has been upgraded from 2.5 to 2.7, this patch does not include it. Applying the patch directly may cause CUDA acceleration to fail. It is recommended to download the full package to update.

huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-PatchUpdate-3.83.7z?download=true

Baidu Network Disk: https://pan.baidu.com/s/1KEfJnoJp8Vavj_T3_jxs1w?pwd=anf7

The pre-packaged version is only for Windows 10/11. For macOS and Linux, please deploy from source
To upgrade a source deployment: pull the source code again to overwrite, then execute pip3 install --upgrade torch openai google-genai whisper elevenlabs gradio_client faster-whisper

v3.82 Full Package Download (2.5G)

If you haven't installed a previous version, please download the full version here. For CUDA acceleration, you need an NVIDIA graphics card with CUDA 12.8 and cuDNN 9 installed.

Baidu Netdisk Download Link: https://pan.baidu.com/s/1whe0SIlx_qK8GV7-mzN9dA?pwd=4q6s

huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-videotrans-v3.82.7z?download=true

v3.82 Patch Package (360MB)

Cannot be used alone; it is only for overwriting an old package. The difference from the full package is that it does not contain the torch module.
Since torch has been upgraded from 2.5 to 2.7, the patch package does not include torch. Directly overwriting with this patch may prevent CUDA acceleration from working. It is recommended to download the full package to overwrite your existing installation.

Baidu Netdisk Download Link: https://pan.baidu.com/s/19juKWCnHMp91bJx5ZtKo1w?pwd=dx7s

huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-PatchUpdate-3.82.7z?download=true

预打包版仅适用于Windows10/11，MacOS和Linux请源码部署
源码部署升级：重新拉取源码覆盖，然后执行 pip3 install --upgrade torch openai google-genai whisper elevenlabs gradio_client faster-whisper

Change

Fix: 修复几处bug 根据bbs反馈
Refacor: 统一报错弹窗
Fix: Google翻译渠道

v3.75以下版本必须下载完整包覆盖，不可仅下载补丁包，因底层torch升级不兼容

v3.80 完整包下载(2.5G)

如果未安装过旧版本，请在此下载完整版，如需cuda加速，需有英伟达显卡并且安装cuda12.x及cudnn9

百度网盘下载地址: https://pan.baidu.com/s/1UiXiX-fGvOoua-s7cMpW5w?pwd=3y6g

huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-videotrans-v3.80.7z?download=true

v3.80 补丁包(360MB)

不可单独使用，仅用于覆盖旧包，与完整包区别是不含有torch模块
因torch从 2.5 升级到 2.7，补丁包不含有torch，直接下载覆盖补丁包可能导致无法cuda加速，建议下载完整包覆盖

百度网盘下载地址: https://pan.baidu.com/s/1EfLHFse3I9ejfk3gNveshw?pwd=bqfg

huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-PatchUpdate-3.80.7z?download=true

解压注意事项：

解压7z压缩包，解压后，找到 sp.exe 文件，双击即可启动软件。
不要将软件解压到桌面或 C 盘的 Program Files 等需要管理员权限的文件夹下。
避免路径错误：解压路径中不要包含中文、空格或特殊符号。
如果使用7zip软件解压提示错误，可尝试使用360压缩解压(https://yasuo.360.cn) / 注意避免全家桶

预打包版仅适用于Windows10/11，MacOS和Linux请源码部署
源码部署升级：重新拉取源码覆盖，然后执行 pip3 install --upgrade torch openai google-genai whisper elevenlabs gradio_client faster-whisper