AI_Girlfriend icon indicating copy to clipboard operation
AI_Girlfriend copied to clipboard

Published 20 hours ago • verified publisher icon gmongaras

i have problem with Initializing custom image movement module

Open tomik520i opened this issue 2 years ago • 15 comments

obrázek_2023-08-08_152035527 can help me?

tomik520i avatar Aug 08 '23 13:08 tomik520i

It looks like pickle is having issues loading in one of the pretrained image mover models. I wonder if this is a torch versioning issue. What version of PyTorch do you have installed?

gmongaras avatar Aug 08 '23 16:08 gmongaras

i complete reinstall all on newest version and have still same problem obrázek_2023-08-08_191318124 i have this version now obrázek_2023-08-08_191429647

tomik520i avatar Aug 08 '23 17:08 tomik520i

During development, I used torch version 1.0. PyTorch probably updated the model loader in torch version 2.0, maybe try downgrading to torch version 1.0 and see if that fixes it. I'm guessing it's a legacy loader issue, but could be mistaken.

gmongaras avatar Aug 08 '23 17:08 gmongaras

i must use 1.10.0+ version pytorch obrázek_2023-08-08_222013205 because aceelerator is incompatible with a lower version and automatically overwrites the torche version to 2.0.1 or the latest version and the problem still remains image

tomik520i avatar Aug 08 '23 20:08 tomik520i

Could be due to an incomplete model download: https://github.com/svip-lab/PlanarReconstruction/issues/28. Try downloading the pretrained models from https://github.com/gmongaras/AI_Girlfriend/tree/master/Img_Mover/pretrained using LFS again.

gmongaras avatar Aug 08 '23 20:08 gmongaras

YES! complete! obrázek_2023-08-09_013048366 I had to download the files manually, this is new files image vs files in rar image idk why, thanks man!

tomik520i avatar Aug 08 '23 23:08 tomik520i

Nice!! Glad you got it working. I know LFS is a bit weird sometimes.

gmongaras avatar Aug 09 '23 01:08 gmongaras

Can I change voice model?

tomik520i avatar Aug 09 '23 02:08 tomik520i

I initially had a feature where you could change the voice model given sample audio, but since the custom voice model was way too large and way too slow, I scrapped it. You can still find it in depreciated/Audio_Generation/Generation_Scripts /generation.py if you want to try it. Not sure if the script still works, but it's a good start. I have a few random audio samples I was trying out.

gmongaras avatar Aug 09 '23 03:08 gmongaras

I'm totally noob in this and programming 😅 I need little help with tutorial 😅

tomik520i avatar Aug 10 '23 09:08 tomik520i

Sorry, but at the moment I don't have too much time to work on getting the custom voice module to work. The main issue is it was very experimental and will take a lot of work to get into the main program. From my memory, the result from the custom voice model was ok and wasn't too good. So, may not be worth trying to put in the main program anyways.

gmongaras avatar Aug 10 '23 17:08 gmongaras

can I somehow adjust that picture little self moves?

tomik520i avatar Aug 14 '23 17:08 tomik520i

Sorry, but I'm not quite sure what you mean by this.

gmongaras avatar Aug 14 '23 20:08 gmongaras

To move around a little bit on her own. Sorry I'm using deepl for translating, I'm not good in English, I'm from czech republic 😅

tomik520i avatar Aug 14 '23 21:08 tomik520i

Aside from blinking and lip syncing, I didn't add any other movement to the image. It would be a cool feature to add though!

gmongaras avatar Aug 15 '23 15:08 gmongaras