RDBT | Krea2
Improve aesthetic, details. More natural, less slop-ish. Supports both photograph and anime art.
Dataset contains ~4k images. All handpicked. Zero shiny plastic glossy AI image. All captions are NL from LLM.
Update log:
v1.0: All-in, full dataset.
v0.2: New method. Refined dataset and captions. x2 more images. Also a professional photograph enhancer. (If you start to rent a h200 node, why not dump more data, as it's already as expensive as f*** ...)
v0.1: 1k images, 12 epochs, all default settings, testing.
Sharing merges using this model is not allowed. This is a free model and it will always free. This "restriction" won't affect anyone, it's only aimed at those who steal others' models to sell.
FYI: all my models have special trigger words and latent watermark.
Description
FAQ
Comments (2)
Krea version could use some improvements in the crowded scenes. I tried to generated a person standing in the crown from a high angle perspective in Edwardian setting, and they always ended up much bigger in scale than surrounding pedestrians. Also, "arms behind the back" pose is completely ignored while the int8 krea2 can show it no problem.
i wish uncensored anime ver existed