The idea was that the model would look like graffiti on a wall. As a result, it was not possible to achieve the desired effect, but it mixes bright colors well. So it's failed LoRa, but may be it can help somebody make cool pictures. The training took place using 50 pictures, stable diffusion 1.5 in this site. I'd recommended use it with anime, cartoon and semi-realistic models. The best result showed portrait view (upper body and close up), cause training was on portrait picture.
Weights: 0.8-1
Sampler: euler a
Hires.fix: i usually use AnimeSharp with anime and cartoon style, esrgan with semi-realistic and realistic. But you can experiment.
Well, ut's my first public LoRa, so if you've needed any more information, just ask)))
I hope that you'd be satisfied with results)
