The purpose of this LoRA is to produce side-by-side images of a subject, one wearing clothes and the other nude.
Since the text encoder is not trained, it may need a little push to get it to provide the correct composition. Try prefacing your prompt with these instructions: side-by-side clothed and nude images of [a woman] wearing [clothing] replacing [a woman] and [clothing] with whatever particulars are relevant to your composition.
Version 4.9 - trained at 1024x1024 and finely tuned for better photorealistic female anatomy. Not as flexible for male subjects and illustrated/animated styles.
Version 3.9 - trained at 1024x1024 with limited synthetic data. Slightly biased toward occasionally generating blurry/lower quality photographs
Version 2.5 - trained at 640x640 - Produces better results than v1. Somewhat biased toward synthetic data with unrealistically smooth skin.
Version 1.0 - trained at 512x512
Description
Initial version for Flux, composition requires explicit instructions in the prompt
FAQ
Comments (2)
Why women look like they are from SDXL generation? Did training dataset contained AI generated images mainly from SDXL and no real photos? Flux can generate higher quality women.
This is very impressive! Possibly a dumb question – would it be difficult to create a LoRA similar to this that has three panels side by side, with the person progressively taking off their clothes? (So clothed-->underwear-->naked)