Lora to generate a view of a women through a window. It's possible to do without this lora, tbh, though it helps. Have a play, and see what you get.
using the trigger word: window_voyeur, will generate a nudey-type view, so maybe that's what you're looking for? avoid that and the images will be more sfw.
Description
FAQ
Comments (8)
What if there are multiple windows, like 4 or 6? Using the original Z-model, I found it couldn't properly control the proportions of the characters within the windows, often generating one window with a huge proportion of characters and another with a normal proportion of characters.
Unfortunately the lora wont help much. I've spent some time trying different things and from what I can see there're hard limitations that make it difficult - understanding of building structure, understanding of the 'view inside a room' vs 'view from outside' and the proportions of things within the view, the 'view port' of the window vs the presence of the window structure, the size of the image generation giving the model more space to incorporate the prompt whilst given each item enough detail... i think it's too much.
i think realistically, this kind of thing is better achiveed with outpainting
Okay, thank you. PS: The Z-Base model has been open-sourced!
Sort of similar, would you consider a lora for a view from outside a car looking in? I played around with just prompts and the view always seems to end up of the occupants front the inside
Hmmn, sounds like an interesting one, I might try that!
@hamstornado Oh yes please do it.
Can this generate males? Couples?
The training images all had females, so probably not