Stomp

-This is the first version of this model. I have made this lora a few times and this was the best result so far. I will keep trying to make it better but this version does get the concept right most of the time. If you use it with a foot lora at the same time it can help with anatomy problems which seem to be the only issue. When I generated pictures a good picture was made about 50% of the time. It was made with hand written captions on 45 images, it still has a trigger word which was included in each caption. The trigger word is st0mp. I also tried 24 images and 135 images but this one came out the best, one of the best parts of it is if you do not describe what the girl looks like much the variety will be high. All of the training images were resized by me prior to training to 1024H&W. The traing images were not cropped.

-Every Picture shown is a first generation without the use of any other lora at the same time. No detailers, upscales, or fine tuning were used in any of the images shown.

-I have only ever used ComfyUI so you would have to try it in Forge or A1111 yourself. I do not even have Forge or A1111 on my computer.

-Use it on a high strength setting from .8 to 1.2

-Most of the images shown were made with Deis, SGM_Uniform, 2 to 2.5 Guidance, 20 steps.

-Euler also works fine with the same Guidance and steps, I didn't prefer the results with Euler but you might.

-Simple, beta, and normal also work as schedulers but my preferred was SGM.

-I don't like the results using schnell, id suggest only using it on Dev based checkpoints.

-Some examples of prompts used were:

- The image is a high-resolution photograph taken from an extremely low angle, emphasizing the perspective of looking up at a young woman's detailed foot from underneath the foot. The focus is sharply on the sole of a woman's foot, which is prominently displayed in the foreground. The foot is bare and has a natural, smooth texture with slightly pinkish undertones. The young Woman's body and face is also visible, face looking directly at camera with mix of amusement and curiosity, st0mp.

-The image is a photograph taken from an extremely low-angle perspective, looking up at a beautiful woman standing over the camera. One of her feet is prominently displayed in the foreground, with one foot slightly raised above the viewer and the other foot flat on the floor out of view, giving a close-up view of the one sole that is right above the viewer. The skin on her foot appears smooth with visible detail and slight creases, typical of natural human skin. Second foot completely out of the image, Ultra Quality, st0mp.

-I'd recommend using the following things in your prompts such as: extreme low angle, standing up over the viewer, or worm's eye view. Trigger word: st0mp

Those three work well in your prompts. It is not perfect and I will continue to try and make it better.

-For ComfyUI users you can really force different perspectives with this node and it looks cool. Although I did not use this node in any of the example images. In the manager just search up: Flux Prompt Generator, author is fairy root

https://github.com/fairy-root/Flux-Prompt-Generator?tab=readme-ov-file

-I don't care about the file size, i care about quality and out of 14 attempts, this was the best one.