A versatile and artistic Illustrious model merge family primarily based on ZU_KI's fantastic ZUKI anime ILL model. As always, this merge is tuned to my personal aesthetic preferences, and simply happens to give me the best (in my opinion) outputs from an Illustrious-based model architecture.
Description
Just some more refinements to the style.
FAQ
Comments (9)
better than v7?
Can you enable it for the on-site generation please?
All of my models have that option enabled from my end, and recently, the staff had enabled generation across the site. However, they just disabled it again yesterday, so it's returned to popular models only. Mine are, unfortunately, not popular enough for them to enable on-site generation for. If it was something I could make happen from my end, I would.
@AI_nstein Thank you for the replying
this isn't noobai?
I don't believe I've used any NoobAI in this merge; if it's in there, it was in one of the models I originally based the series on (Zu_ki's, to be precise.)
I personally don't like the results with NoobAI, so I've been avoiding adding it to my merges.
@AI_nstein i see, because my lora that trained on noobai worked well in previous version of this model (to be precise v6c and older), unlike other illustrious model merges. maybe now in v7 and v8 there are less noobai influence in this merges
@PertaliteMeister I won't pretend to understand what really goes on under the hood in the merge process, so I'm not qualified to speculate on why that is, exactly. However, I did do a slight cross-merge with my Custom series (which is, in turn, a minor cross-merge with my XL/Pony series) in an effort to make it more unique, and that may also have contributed to polluting data the LORAs would need.
All of this is extremely experimental, and the result of my own ongoing efforts to simply test the limits of the technology, and I have to confess that after almost three years of such experiments, I feel like I still barely understand any of it sometimes.
@AI_nstein I admire your attitude to this project. Thank you very much for the checkpoint.












