Introduction

Welcome to my newest model, where i stand on the shoulders of giants. I merged the models Hyphoria by Ecaj and Hassaku XL by Ikena and added in the workflow for the Rouwei Gemma Adapter by Minthybasis.

What i tried to achieve is a model which supports MANY artist tags (thanks to Hassaku), has superior image quality (thanks to Hyphoria) and superior prompt understanding (thanks to the rouwei gemma adapter). I believe LLM prompting support for illustrious is th way forward.

No lora, only prompt

All the images i have posted here are made without lora, just pure prompts and artist tags. See my article on all the things i did in my workflow if you want to understand how i made these images possible. Or just grab my workflow from model post or article, install the nodes and go!

(careful, the extra images i'm posting below have an older version of the workflow. it's probably a lot more messy)

Artist previews

Inside the training data file you'll find an artist preview collection. The collection is not complete, and will be updated once i have more artists.
Artists are rated by score, 0 being the reference image, as if no artist would've been used. 20 being the most different. Included is a mini gallery browser which launches a python webserver. On the top of the included gallery are a Min and Max score rating slider. use those to remove results of artist too different to default, or those which significantly change the image. My recommended values would be to set Min to at least 07.

Need help?

I can often be found lurking in the CivitAI discord chat with the same name Eface. Don't be discouraged to show me your whackiest creations!