The Sundress.
Can it take over the world? Maybe not.
Should it, though?
That's the real question, isn't it?
This here is my first public LoRA and it's specifically generated for sundresses of all kinds. Trained on 156 entirely SFW images and the epiCRealism XL Pure Fix base model, this will maybe be your one stop shop for all things sundress.
It works comfortably around 0.5 - 0.8 (0.55 being my favorite) and shouldn't get in the way of identity LoRAs.
If you love it, let me know! If you hate it, constructive criticism is encouraged since I really have no idea what I'm doing.
Enjoy!
Description
Version 3.6:
The Sundress LoRA was my first and needed some TLC. Here're the main changes --
Stripped out any headless images from the main dataset. Don't want to pull training toward decapitation.
Added additional 3/4 and full body images to refill the dataset.
Initial caption files were a mess. Those have been cleaned up, tightened, and normalized.
Ran a de-bias phase to steer away from default accessories (hats, sunglasses, handbags). Should be less common now and easier to stop with negative prompts.
Re-balanced tag distribution.