Jandown

 之前也是花了大量时间没搞定，不甘心，又花了两天时间，终于成功了，结果卡在step2了，大概率是显存不够了 估计也只有3090/4090能跑的动了，其他卡就别浪费时间了

pip install transformers==4.39.3 这是transformers版本不对导致的，我是2.1.2的torch，安装对应的transformers就可以了

> Thank you. It worked. So it uses Japanese audio for English song. Did you figured how to use different voices. I also encountered the same problem and did the...

> 看下环境里magic-pdf的版本号是多少 0.10.5

> 可以尝试降级到0.10.2再试试，目前在0.10.3~0.10.5发现一个性能bug，已经安排排期修复了 感谢解答，尝试降级0.10.2还是报错，前两个样例用doclayout_yolo模型正常，用layoutlmv3则报错，第三个开始，两个模型都是报错 File "E:\AI\MinerU\app.py", line 114, in to_markdown local_md_dir, file_name = parse_pdf(file_path, './output', end_pages - 1, is_ocr, TypeError: cannot unpack non-iterable NoneType object

> 你看看堆栈里是不是file_path为None，如果出现这种情况，一般是上传新的pdf时，gradio没有处理完上传，也没更新file_path时就点了convert导致的，可以尝试每次新上传之前点一下clear按钮，或者上传pdf后等待PDF preview窗口可以正常渲染pdf之后再点convert按钮 应该不是这个问题，我首次启动webui转换就报错了，刚试了clear，也是一样的错误。

> Any context? What do you mean changing the video?任何背景？你是什么意思 更改视频 What I mean is video face swapping. The prompt shows that the face change was successful, but the video...

> Ok I might have found the solution. You need the FP32 model of the inswapper > > `https://huggingface.co/hacksider/deep-live-cam/resolve/main/inswapper_128.onnx?download=true` This problem is like a ghost, often appearing. Today we updated...

> > > Ok I might have found the solution. You need the FP32 model of the inswapper > > > `https://huggingface.co/hacksider/deep-live-cam/resolve/main/inswapper_128.onnx?download=true` > > > > > > This problem...

> > > > > Ok I might have found the solution. You need the FP32 model of the inswapper > > > > > `https://huggingface.co/hacksider/deep-live-cam/resolve/main/inswapper_128.onnx?download=true` > > > >...