Howdy yall if any of you have been looking have fun with ai i've been cooking up a proof-of concept LoRa to see if complex words/phrases/symbols can be generated consistently by stable diffusion (particularly in the context of "bleaching" an image using in-painting)
here is an example of an image that turned out well
https://imgchest.com/p/6eyrgrbel4p
anyways here are the things your gonna need
https://www.dropbox.com/scl/fi/vg424jkmagx75powe3dq1/WCM-10v2.safetensors?rlkey=1fvxm7gnzdu251rzfq0vkx81a&dl=0
also the following resources are REQUIRED: https://github.com/AUTOMATIC1111/stable-diffusion-webui , https://www.python.org/downloads/ , https://civitai.com/models/165985/anylora-checkpoint-fp16cleaned as well as a recent-ish NVidia compatable GPU with over 4 gigs VRam and 10 gigs hard drive space with an install of Linux or Windows 7/8/10/10 pro/11
to use the lora simply put it in <stable diffusion directory>/models/Lora and add 'white cocks matter shirt <lora:WCM-10v2:1>, clothes writing, english text, text on clothing' to the positive prompt
note: AI is prone to scew ups so if it doesnt work at first try generating again a couple more times and if it still fails adjust positive and negative prompts as needed
you should leave model weight at 1 (the 1 in <lora:WCM-10v2:1>) for best results
please note that is HIGHLY experimental and even though I will be gradually releasing more LoRas I have limits: no overly complex topics, and keep things on topic to the premise of the bleached fetish/kink