If you liked the model, support the dataset images creator.

25/11/27 V1

40 images 1024x1024 on basic IllustrousXL. I've been experimenting and one of the insights was that you really shouldn't add background tags like "simple background/color background" too much. Model is just confused in result.

The style is not really like the artist's style but still very neat. There was much preprocessing of the images since all the posts were in compressed video format, so I manually cut the frames and cleared the palette - very tedious work I'd say. To trade with effort - many tags were set with autotagger and cleared as my attention allowed.

No trigger words needed this time - enjoy.