!!! READ DESCRIPTION !!!

Artist`s Twitter: LINK

This lora is extremely unstable in use, as it was trained on screenshots of abstract works in the form of GIFs (see the artist`s twitter)

I did not expect the final result to produce anything resembling a normal image. In the end, I found the result very interesting (albeit with a major drawback in the form of particles that cannot be controlled properly)

The author's style cannot be conveyed through static images, and his style itself is quite abstract to reproduce using SD-based models. Therefore, I can only state that this lora is based on his works, but is not even capable of closely imitating them

Since the lora itself is not very stable, I recommend trying different weight values (for preview images, I used values from 0.8 to 1.3) and trying different combinations of tags.

The tags specified in the lora details are included in the dataset tags, although some of them are not very stable or require weight increase (“blue theme”, “red theme”, etc.). Tags such as “Absurdism” and ‘Minimalism’ were not included in the dataset tags, but they have an extremely strong and “interesting” effect on the final result