A model based on shiin artstyle.
Description
Tips for use:
•lora seems to handle short prompts better.
•"comic" and "cartoon" help maintain fidelity to the style. •the model is still limited, it is possible to use it together with the previous version (with a weaker strength, of course).
•If you're going to use schizophrenic negative prompts, use SEVERAL. Putting just "bad anatomy" and "bad face" makes it even worse, but a long string of several known tags yields better results.
•"score_8_up" tends to stray too far from style, if you want fidelity use "score_7_up" or "score_6_up".
•use "(ljn)" as the first tag.
*edit: FINALLY, some progress on this Lora. This version is definitely more limited in certain points, but absurdly more consistent in terms of style than the previous one. It's still not the definitive version, but it generates objectively better results and without the need for extremely restricted prompts to keep the style (the previous version looked more like a concept Lora capable of generating wide @sses than a Lora for style, lol). Anyway, if I don't mess around for too long I can make version 2.0 quite robust and without the grotesque defects in the anatomy and faces (GOD, PLEASE SPEED UP THE PRODUCTION OF PONY-V7)
FAQ
Comments (3)
Man your model is goated, I love every image I was able to make with it
I'm great that you liked it, but it's a fact that the training didn't exactly teach the exaggerated anatomy. I hope that with the arrival of ponyv7 and the illustrious models I can train the definitive version. I suggest you test the "Shiinv3" model. My attempt to imitate more...aggressively, let's say, Shiin's more detailed style.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.