Well kinda does what it says.Aspect ratio should be portrait mode to cover the body, landscape won't work well.First time I train a lora where the training data is purely synthetic, because shooting data like this yourself would break a feature film budget in terms of shooting models in all major cities at different time of day in different clothes.Trained on renders done with Flux Ultra, since it handled this kind of shot better than dev.

Files with SHA256: ea797a0c0aeb4d7d233bc570523a5b2eeec11a99a687c874c604d4cb53a89728