V2 (Real Dream)This isn't much of a revision dataset wise, as most of the images are real pictures. I swapped a few of the synthetic ones (there's only about 5). But the big difference here is real dream - 85 training images, 40 epochs (slightly longer), trained on real dream which should give a boost to quality of output. Recommend not using full weight due to those few synthetic inputs.Same as before, main prompt is "POV_Devils3way, mmf, 2boys, 1girl, mmf threesome"Note on negative prompting:This time I added the tag 'semirealistic' to all the synthetic pictures, so if you want to avoid that look, which can be pronounced, put that in the negative tags at whatever weight, and probably if you are getting those you can reduce the cfg or lora weight. Can try something like (cartoon, illustration, semirealistic, sweaty) at some weight. This seems to work!Double penetration will only work with some picture width, like landscape. But does seem to get an "about to penetrate" double penetration result on portrait, which is new. Faces on dp gens do get a bit messy, so face detailer, or a lora is a good idea. Old V1 Trained on a dataset 86 pictures, this is designed for POV MMF threesomes. Could easily train this dataset on real dream and get better results for a v2, because it's largely real data not synthetic - if anyone wants to flick me 1300 credits lol.

POV Devils 3way

This isn't much of a revision dataset wise, as most of the images are real pictures. I swapped a few of the synthetic ones (there's only about 5). But the big difference here is real dream - 85 training images, 40 epochs (slightly longer), trained on real dream which should give a boost to quality of output. Recommend not using full weight due to those few synthetic inputs. Same as before, main prompt is "POV_Devils3way, mmf, 2boys, 1girl, mmf threesome"This time I added the tag 'semirealistic' to all the synthetic pictures, so if you want to avoid that look, which can be pronounced, put that in the negative tags at whatever weight, and probably if you are getting those you can reduce the cfg or lora weight. Can try something like (cartoon, illustration, semirealistic, sweaty) at some weight. If it turns out these synthetic elements remain even when prompted for, I may have to delete them all, and retrain with just real pictures :/

Files with SHA256: 190ca3cfe56ede1f0568c3c232535e439a0ef27ce4dd07c97990fa7f76b66b7a