Important: I haven't had a single incident of Flux turning the small bodies into children, but you it's probably going to happen eventually. So just be aware of this possibility if you are sensitive to this sort of art.
Note: You need to prompt exactly what you want. This is not a Lora where you can just copy in two paragraphs of camera settings and let the Lora do all the work. Also, Flux doesn't want to do any of this, so you may have to take the strength very high (1.6 for no arms or legs, lower if you want them). You will likely not succeed at using a custom face, but there's no harm trying. When you leave off the arms and legs, it is easier for Flux to find room and fit your character into your chosen situation.
Sadly, I don't believe you can successfully add other people to the picture without turning the suject into a normal person.
This is a difficult but, I think, rewarding Lora. My method was to input my prompt, run a few batches, and find the best seed for what I was going for. At least half the time it gave me a good pic immediately, but then I often used that to refine my prompt.
Also, sorry for the bad trigger word. Had I any hope this was going to work, I'd have come up with something better
Trigger: quadampmidg
Description
FAQ
Comments (2)
Holy shit, I’m already celebrating just the sheer brilliance of coming up with such an idea! The actual execution and training of the LORA, though… well, that’s a whole different beast.
As you already pointed out in the description above, "Flux doesn’t exactly do certain things voluntarily." But the fact that—with an excessive amount of friendly persuasion (haha)—you can actually convince it to comply? That’s nothing short of impressive!
So yeah, I’m loving this. And the results? Well… let’s just say: Among other things, they are very spectacular.
Thanks! I just now became capable of doing training on my own computer. This was done with the site standard 20 images, so my plan is to see what happens if I expand the training set considerably.