This is my first attempt at creating a Checkpoint from scratch (albeit trained against Stable Diffusion 1.5). I wanted to create a model that would re-create the British style of "Reader's Wives" amature glamour modelling of the 80's and 90's, in magazines like Razzle and Escort. It does't require any keywords as such, but "large breasts" and "Natural Pubic Hair" will pull out of the LoRA's I trained and one or two others from the community. Feedback appreciated.
Description
This is my first attempt at creating a Checkpoint from scratch (albeit trained against Stable Diffusion 1.5). I wanted to create a model that would re-create the British style of "Reader's Wives" amature glamour modelling of the 80's and 90's, in magazines like Razzle and Escort. It does't require any keywords as such, but "large breasts" and "Natural Pubic Hair" will pull out of the LoRA's I trained and one or two others from the community. Feedback appreciated.
FAQ
Comments (9)
great work, not my taste but its important that every tool is out there
Was the training data only people, or did it include British style environment and settings not featuring nude women? How many images was this trained on?
It's difficult to say how many images because it's multiple LoRA's merged which I trained myself but experimented along the way with different weights. The next version will focus more on BBW. British cities are already in Stable Diffusion. If you try prompts like "standing in Newcastle" or "on a Manchester Street" it gives reasonably British looking environments.
I saw the pic without reading the title. I was like, "WTF that looks like a british 80s house wife pin-up girl". I read the title. You're spot on with this. It's totally not my taste, but I'll be damned if you didn't create an accurate model.
Hi - I like the look of this model and got a couple of good images I was happy with initially. Now it just seems to want to decapitate every woman. Even when I copy the sample image data, chops tops of heads off above the mouth. I'll try more settings, but can't seem to get a head!
You'll need to lower the weights and probably use ControlNet as well.
This model is very to my taste, but I have the hardest time prompting it. The slightest deviation from any of the examples, and I get hardly anything, it seems. Breasts out of context, women without nipples, headless women, etc. Its very touchy.
Send a few examples and I'll have a look at tuning it.
nice!