*Version 05*Use ([white woman::0.xx], [asian woman::0.xx], [indian woman::0.xx], [black woman::0.xx]:1.0) syntax for this versionThis LoRA was trained with a mixture of white, black, and asian references in an effort to make a mixed race model with a bit of play to it. My best results have come from keeping the CFG around 4, the steps close to 20, and the weight of the LoRA itself around 0.5.To tweak the weight of each race further (admittedly by default the model leans towards white the most even though that's the race I used the least in training), use the following syntax in your prompt:"photo of nude ([white::0.XX], [asian::0.XX], [black::0.XX] woman:1.0)"Replacing "0.XX" with anything between 0.01 and 0.99; the closer to 0.99 means the more steps to which that race will be applied.(If you are unaware, the [prompt::0.XX] syntax means "prompt" will be applied to XX% of your steps. So, for example, [blonde::0.78] and using 100 steps means the prompt "blonde" would be applied to your gen for the first 78 steps then disregarded for the final 22 steps.)

MeltingPot

New sample imagesAdded Indian subjectsUse ([white woman::0.xx], [asian woman::0.xx], [indian woman::0.xx], [black woman::0.xx]:1.0) for this version instead of just white/asian/black

Files with SHA256: 889380bc1dc4b8e806d3026733b2a055f65b3ee69bfa14f28f709dd2344360c9