Dzenrei Style Lora.
Two versions:
1) Trained on Pony - used for Pony and it's merges;
(Example images with this.)
2) Trained on Illustrious - This one is to be used for Illustrious, but also works on NoobAI.
(for the example images of IL version, since, again, I have no idea why this site no longer recognizes what checkpoint I've used on an external generator - first and second are with NoobAI1.0, third and fourth are on Illustrious-XL)
Now, three things:
1) I am yet to figure out what are the best options to train a LoRA for NoobAI. Should the base be Illustrious or should it be Noob itself? Any advice is welcome.
2) I still find Pony better and would continue to train on it. That might change in time, mostly after I have enough information and experience with Illustrious and Noob.
3) I will most likely not make IL versions for all my uploaded models. That is just too much work, too much time, too much space on my PC, too much resources full stop. So if I make an Illustrious/Noob versions it would be on select few.
*suggestions and feedback on which ones those could/should be are welcome.
Description
FAQ
Comments (8)
Compared to pony, illustrious XL looks way worse, I don't really get all the hype around illustriousXL and i am speaking in general, i just find the new base model underwhelming
I think so too. But I haven't tested enough of Illustrious and NoobXL. I definitely will do some testing, but when I have a bit more free time. Until then Pony remains as my main and go-to for both genning and training.
Same opinion guys. Illustrious looks good here. But when i make gens,the result is not impressive, compared to Pony.
@mikebobby681369 what i found out is that you gotta use a lot of "masterpiece" and "high quality" type prompt to up the quality, literally changes the image from amateur dog water art to something half decent, its simple but very primitive to pony's score system
@ninjahatoori8; @mikebobby681369 Here's my experience so far: all the gens I've done with both Illustrious and Noob (but this applies mostly to Illustrious) have come out oversaturated, lacking any shading and/or highlights, and as bonus, the body proportions are a bit too slender and childish (anime style, yes I get it, but come on...).
New models will always have hype around them. Flux was the same. I still think its hype and merits are even more overblown than IL. Glorified, cleft-chinned Midjourney copy, but THE WOOORDS, man, the woooords! Because image generation's main purpose and thus incentive was and always will be word creation :D
Illustrious models don't seem to work with accelerators like lightning or hyper so that makes it an automatic no go for me.
Awesome style!
Indeed. I really like it, and it's a bit of a shame the artist has so few works.