V2: Recommended settings are with DMD2. Secondly CFG is real touchy with this model. Better most of the time at just 1. 1.5 if it's hard to land.
https://tensor.art/models/806617089210062925
Generally does crown braids and Victorian hair pretty well, as well as lattice windows, stone walls, Victorian floral wallpaper etc, but you may need either kingdom come, or Victorian SDXL loras at 0.3 or 0.4 or so weight to sink some things (I have both reposted). Can prompt with victorian style hair, crown braid, medieval, Victorian, old west, pompadour etc.
Sometimes the half trained lora is better at Victorian hair. If you want to nail updo's that have ringlets rather than are just frizzy, give that a go. It'll do some other things less well, like setting and clothes. But for the Victorian hair it's often better, hence why I left it up (V2 Light)
Cyberpunk limbs will also work most of the time, but background/setting you'll have to prompt carefully. Ofc, as a nude lora, I didn't upload a lot of backgrounds, so that's broadly true - complex backgrounds might take a lot of fiddling to gen right.
This is CHEAP AS HECK now to generate with, so that it requires a bit of regening is no biggie. Knock yourself out. The non-solo pictures took me a while of fiddling to get to come out, but just to show it can be done - fairly period accurate stuff (mostly).
Generally real happy with this!
V1
This main model is tuned on the lustify (endgame) sdxl model, trained on a dataset of artistic and tasteful nudes. That turned out perfect (lustify endgame is a great base for this) and is honest some of the best quality nudes you can generate IMO. For some contemporary pictures, I think this can be better, it's trained a bit heavier, but will depend on the specific picture.
Flux however, is a bit of a mixed bag. Note the flux version is fairly experimental and does not take instructions or generate coherent pictures nearly as well. So for flux variants: Maybe 0.4-0.5. Just to reduce the weirdness.