Big Booty Bikini Woman On The Pool Divider (Illustrious)
Details
Download Files
About this version
Model description
At high strengths, this model makes a very specific pose in a specific scene. It's a woman in a bikini with a large booty sitting on a divider between two pools with stone tiles. It started from a single image but I developed a whole dataset from it, mostly adding variations so it'd be less opinionated about elements such as clothing, hair, style, etc. You should be able to substitute whatever character you want into this scene using this LoRA. Have fun!
Alternatively, you can make some variant scenes using lower strengths. It can help with things like having a character's feet in the water.
This one requires several captions to function properly, but pool is the most important trigger word. from above, sitting, ass are almost as important. hand on thigh, arm up control arm positioning and are needed for the intended pose. So the recommended minimum captioning for the intended pose is: pool, from above, sitting, ass, hand on thigh, arm up
Medium captioning is: woman, medium breasts, brown hair, red bikini, thong, pool, sitting, hand on thigh, arm up, feet in water, from above, ass
The full captioning is: woman, medium breasts, brown hair, ass, red bikini, thong, outdoors, pool, sitting, stone tile, looking at viewer, back, full body, bracelet, earring, hand on thigh, arm up, feet in water, from above
The model also knows a few variations on clothing and hair. Longer captions make the effect stronger, and you can use them as negative prompts if you have trouble getting what you want to generate.
By default, if you get it to generate the scene correctly then it skews towards the Gerph style, so if you like the style go give Gerph a look. It's also been regularization trained on Oatd and Galena styles.
If you used the Pony version of this model, make sure to add ass to your captioning as Illustrious needs it to get the pose right. I'm starting with a new model card because I'm rebooting the version numbers to v0.1 due to the need for better regularization data to get the best result.



















