-\/- V2 Description -\/-
This LoRA is designed to create more accurate quicksand scenes and more consistent mud-covered effect. It can work with strength between 1-2 creating okay results.
Mud-Covered: Use 'color mud-covered' optionally with 'mud-covered x' to create the effect. For example: 'brown mud-covered, mud-covered bikini'. The tag 'colored skin' also helps enforce 100% mud-covered if that is what you want to achieve.
Quicksand: Use 'x-deep' paired with a description of the location to create the intended effect. For example: 'waist deep, forest, day'. Tags like 'partially submerged', 'wading', and 'quicksand' can help, but they also have their own strong biases that can introduce unwanted effects so be careful with them. However, I found that including 'color quicksand' generally didn't cause too many problems and helped control the color of the quicksand well.
These are the specific trigger words trained in the LoRA: 'brown mud-covered, grey mud-covered, green mud-covered, yellow mud-covered, black mud-covered, neck-deep, chest-deep, waist-deep, thigh-deep, ass-deep'
I believe the only way to improve the quality would be for me to find more and higher quality mud and quicksand illustrations. Unfortunately, such illustrations are rare and not easy for me to find. If anyone knows where I could find more, it would be greatly appreciated.
-\/- V1 Description -\/-
This is a LoRA trained on muddy faces, hair and bodies with various levels of sinking into quicksand or mud. What it does is provide extreme mud coverage and color control. Base models give you small mud splatters that are always brown, whereas this can make 100% covered or made of mud body parts and clothing in the color of your choice.
It can generate colors other than trained ones, such as red, blue, pink or yellow.
When making quicksand scenes, try using 'close-up', 'portrait', 'upper body', etc. to help her sink deeper.
If you like what this can do, please take a moment to post an image to this model, it helps me out a lot. Thank you!