I need some advice on AI I have 2 particular applications that I experimented with various degrees of success.

1. Copying someone's face from a batch of their pictures. There is a married chick and i scored some pics of her face on her Instagram, but they are of shitty quality, still I think they are good enough to use that google AI training tool, but each time when i try to use it and generate an image i just get static as output in stable diffusion.

I have also tried to use chatgpt to in super hyper elaborate detail to describe her face as if its a forensic artist, i had like 5 page description of her face and that has given me the best result highest quality result, but the result was not consistent, some pictures looked like her others didnt, and clearly was not her in the image, but best result tbh. (I basically fed that entire description back into ChatGPT and begged to generate an image in either leggings or shiny silky body tight dresses and 50% of the time it would agree and generate).

For google face training thing i had 10 pictures, they were all a bit blurry but using Grok for example to turn picture into video - that was sufficient to make her face move animate etc. so IDK why that google tool thing ate shit.

2. Because 1 ^ did not work, I resorted to using Stablediffusion editing image trick, basically carefully selecting clothes out and gambling with the generations which in some cases were okay and produced decent fapping material, but nothing would work better than having her face and doing anything i wanted with it. Images obviously sucked, were blurry as shitty screenshot or crappy facebook quality pics.

Any tips? I could try AI upscaling some of her photos before training AI that could be a method, but all upscalers ive used suck dick, maybe stable diffusion can upscale, idk..


picrelated is my goal