(Also works for any other type of style transfer, but my initial interest was with realism)

In my time experimenting with realism models I've always been disappointed, as I could never get the sort of creativity, variance, and quality I wanted.

But with Krea 2 I've found something interesting. I'm able to make creative stylized images with a model like Illustrious or Anima, and then run a denoised KSampler over these with Krea 2 to get good realistic-looking results while generally retaining the creative features.

Once you get everything configured right I've found the outputs look even better than what Krea 2 is able to generate through prompting alone.

(It is recommended to use realism LoRAs to help push the effect, as Krea 2 might be insufficient by itself.)

_________

There are two workflow approaches to this:

An "automatic" workflow which uses the vision of the text encoder to inject information about the input image directly into the prompt, providing a quick and a fairly accurate transfer, but also injecting some stylized information, so it's not ideal.

A "manual" prompting workflow that provides stronger and more controlled realistic style transfer, but requires the use of LLMs to help with the prompting.