I've only recently discovered Webui, so I used Civitai to create my first lora.

And I trained LORA using relatively recent paintings by the artist 'youjomodoki' as data but I'm new to this process so, my model might be a bit sloppy.

When creating this lora, I used the SD1.5 base model and the 'CAT - Citron Anime Treasure [Illustrious & SDXL & SD1.5](illustrious)' checkpoint.

+

I'm still testing a lot of things, but I'm using 30 steps with DPM++2M/krass as the sampling type.

If you think Lora's personality is weak in the generated image, you can try adjusting the weight to less than 1 to 1.5.

And it's even better if you use Hires.fix to correct the image you've created.

Good luck!! 😎👍