Page 1 of 1
[Newbie] Convertion Input Video different than extracted/trained
Posted: Thu Sep 07, 2023 2:31 pm
by adriadg
Hi,
If I use a video 1.mp4 for person A for extracting and training, can I use a different video (2.mp4) that appears the same person A for converting?
I mean, should I have to train for each input video that I want to use for conversion?
Or if I have an enough trained model (A&B), can I use this model for other different videos of person A?
Hope you can understand my meaning.
Thankss
Re: [Newbie] Convertion Input Video different than extracted/trained
Posted: Thu Sep 07, 2023 9:26 pm
by torzdf
Yes you can. It works better if you have trained with a variety of sources for A.
Try running a convert and see if you are happy with the results. If not, you can always train your existing model some more with new A data.
Re: [Newbie] Convertion Input Video different than extracted/trained
Posted: Fri Sep 08, 2023 1:15 pm
by adriadg
I have tried to run a convert but an error message appears, something like this:
This message does not appear if I use the original video for convertion.
So, what do I suppose to do now? Extract first the faces from new video of person A for updatting the alignment file? Or what?
Thanks
Re: [Newbie] Convertion Input Video different than extracted/trained
Posted: Sat Sep 09, 2023 9:33 am
by torzdf
adriadg wrote: ↑Fri Sep 08, 2023 1:15 pm
So, what do I suppose to do now? Extract first the faces from new video of person A for updatting the alignment file? Or what?
viewtopic.php?t=1083#Alignments