Two versions, and all round model, and one that specialize is wedgies from behind. The specialization model produces much better results but only does the one thing.
It works and quite well sometimes. Good for in-painting/Sketch too. However finding the sweet spot settings and prompts for each model is a real pain. Prompt words especially are very testy.
My recommendation it to keep CFG low, and Lora strength also low maybe 0.5/0.6 and then work your way up on each model.
For prompting, I'd couple (ass) with back_wedgie and (partially visible vulva) with front_wedgie. But that's about the only general advice I can give.
hanging_wedgie tag was experimental and kind of ruined the training because of all the western images. But I'm not going to retrain because it took far too long. Pretty unreliable but tagging flagpole or tree seems to help.
Description
FAQ
Comments (6)
Very nice concept. I did try to do the wedgie LoRA thing but it's so complicated (and not so many images to train with). Is this LoRA going to be improved? I'll try some generations and post my best results.
I doubt I would revisit this one. Training times were too long. I do have a more focused 'from behind' version that produces better quality stuff I might upload later.
@fongletto please do, any additional wedgie AI tools would be wonderful
@dokudoku I uploaded one that specializes in wedgies from behind. It should provide better results but I never fully tested it.
@fongletto Thanks, indeed, is more limited but works much better!
@fongletto Thanks so much!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.