jcarlosd32

/

mrdeoro-lora

like 0

Thanks a lot John, your fix worked and now I can use the model as I wanted. I have valued your response in the forum, thanks again.

I will open a different question because I trained the Lora model to generate pics of myself, and it is working great when being alone in the photo. But when adding other characters to the scene, everybody looks like me as well. I do not find the way to specify it in the prompt, so I can not generate scenes with more actors than just myself. Maybe this goes beyond your knowledge....but just in case.

Gracias!!!

If it works, it's OK!

I kinda get what you're saying...
It's a LoRA strength issue, sometimes called scale or weight.
For example, in multimodalart's space, you should be able to specify it. I guess it's not possible with the server settings, it's fixed at 1.0.

Anything more is beyond the scope of my knowledge. That sounds great, but I am simply not a LoRA producer.
Maybe there's a wiki or something with the know-how?
I guess it's not absolutely impossible, since there are generated illustrations with multiple animated characters. (I hear some people generate images in segments and combine them)

Indeed, it should be possible, I saw plenty of examples where images like "a photo of xxxx (being xxxx the person used to train the lora model) and superman/batman/whatever was produced smoothly. Or situations where xxxx is surrounded of others, and they do not look all like him. In my case it happens even with female characters, they all adapt to look like me hahaha

When I try to adapt the Lora strenght to values lower than 1, then everybody in the picture is less similar to me, so it applies globally....and does not solve it.

I will look in forums or post a different question. Thanks again for your support and comments!!

Yeah. Ask someone who knows! That's the best.