Who needs style LoRAs? 😉

Illustrious is very powerful when it comes to styles. Its default Anime style is just a small part of what it can do.

So before you slap on another LoRA, give the base model a try. Odds are, it already knows your favourite style or artist.

Here are two wildcards I have been experimenting with, and I hope they can be an inspiration for you to build on and push further.

I suggest to try them with minimal positive and negative quality tags, especially for the more unconventional styles. In these images I've only used

masterpiece, very aesthetic, score_7_up

as positive, and

score_1, score_2, score_3, artist tag, censored

as negative base tags.

All images show the styles at 1.1 strength to emphasize the effect, but you can go either way of course.

illu_lora.txt contains a few artists and styles, including additional prompts to anchor, but Illu also works fine with just the artist.

In illu_lora_polar.txt I've been experimenting with narrowing down styles using both positive and negative prompts, which grants infinitely more possibilities. This is more like a proof of concept for me, but fun nonetheless.

In ComfyUI, you would need the "Negative Wildcard Processor Node for ComfyUI" which moves our <!...!> prompts from the positive to the negative prompt. Alternatively, you can always just move them over by hand.

Give them a try, show us what you created, and have fun ❤️

And if you know more nice style/artist collections, let me know!