안준영 (Justin)

Thanks for your quick help!!

In addition, I just checked variables for paths. They look invalid in my opinion.

> You can use the old release version of SadTalker [v0.0.1](https://github.com/OpenTalker/SadTalker/releases/tag/v0.0.1), and change the checkpoint_dir = 'checkpoints/' Thank you so much, but seemingly SadTalker v0.0.2 and v0.0.1 have different directory...

>  Here is my folder architecture. I created a new directory "src" and copied some files into it. You can also delete duplicate files. The "utils" is copied from...

> The sadtalker I used is v.0.0.1. The files in the folder "src" are copied from v.0.0.1, except for "utils". The v.0.0.1 doesn't contain "utils", so "utils" is copied form...

> The sadtalker I used is v.0.0.1. The files in the folder "src" are copied from v.0.0.1, except for "utils". The v.0.0.1 doesn't contain "utils", so "utils" is copied form...

> Glad to hear that. > I fed audio and a source image, and the generated talking head results are fine. That's interesting. Do you mind sharing the code (the...

Same question here. Any progress?

@huangwei2913 Hello, thanks for your information. So you mean first we need to download 24G dataset then extract some of them according to those numbers?

Seemingly the output is only resized, not aligned.