Faceswap Forum

Hi everyone,

first of all thanks for this useful forum pages and for the whole Faceswap endeavour.

I am writing looking for some help.

I am experiencing this problem right now.

After being successfully able to convert one mp4 video into an other, I am now stuck into something I am not really able to understand. Using exactly the same settings, and after a successful training time, the convert process does not work.

The event windows keeps saying (for every single png): "no alignment found for [], skipping".

The first time I did it it was about a 63 Mb mp4 480p video file. Now I am using a video that I have personally recorded of about 2.8 Gb mp4 1080p video file. May this be an issue? I am trying to resize it to a smaller size.

Thanks for your help.

Alex

It's saying that it doesn't have alignments for that file. Did you run the extract stage on that file and if so did you select the correct alignments file for the convert task?

In fact I was waiting for this first post to be approved in order to give an update. To answer to your question I have discovered that it automatically spots the alignment file if not inserted. And that I found to be true. Talking about the size of the video in fact I reduced the video from 1080 (height) to 480 and it actually worked. So: case solved
Thanks!

Size of the input video shouldn't make a difference.

Do you mean that it does not matter for extraction? Or you mean it does not matter for conversion too?

If for the conversion the size does not matter than I shall figure out why changing size (1080 to 480) makes the conversion process work.

If you change the resolution between extract and convert, then the process will not be able to find the faces, as the hash will have changed, so the alignments file will no longer represent the original video.

The resolution you extract at (and subsequently convert at) has no direct impact on the ability to find alignments.