Update Sept 2025: Due to civits unfortunate deplatforming and downspiral into insolvency i will no longer be able to keep this model on the onsite generator. Have fun with WAI or the 15 versions of Nova NSFW 3d furry or whatever.
I wanted something that I could consistently use to reference baked-in illustrious artists, along with my style and clothing loras examples, with a slightly more aggressive photorealistic bias. It wont win any awards for photo or realistic but it'll get you drunk (metaphorically?).
For those onsite gen only, most of the original models I would use as reference are no longer available on any generator anyways due to Civit's 200-slot system and the need for mergers and trainers to chase novelty to stay relevant.
I don't plan on updating this one as I would prefer to work with loras on a consistent foundation.
To the best of my knowledge, there is no NoobAI in this, eps or otherwise.
I typically use the following bog-standard pos:
masterpiece, best quality, very aesthetic, newest
Neg:
3d, (blurry), child, loli, aged down, (lowres:1.2), (worst quality), (low quality), (bad anatomy), bad hands, multiple views, jpeg artifacts, signature, watermark, text, logo, artist name, (patreon logo patreon username)
Description
SELECT A VAE WHEN GENERATING ON-SITE IF YOU DONT SEE ONE
FAQ
Comments (14)
Hey hey hey, wait a second!
"I wanted something that I could consistently use to reference baked-in illustrious artists, along with my style and clothing loras"
Does this mean all your style and clothes loras are included? like, all of them!?
<insert xmas came early.jpg here>
If yes. Amigo, you just made my day. If no: Amigo, get to work pronto.
(what, y'think you can be the best clothes Lora creator on the entire site, and just rest on your laurels? your life belongs to us now...)
Unfortunately I just mean a model I can use as my reference model. I tend to bounce around with models in the demos and all the original models I use have like 8 versions, so it avoids confusion if there is just one model I post to.
I dont think baking in all the clothing loras will work very well, it doesn't work like that..
Thanks for the kind words though.
@Sophorium Ah well, that's life. We struggle on nevertheless!
Your welcome (your outfit loras in particular are consistently of superior quality, as compared to others in the site's top creators). Keep it up
(one glorious day in the far future, all our names will be emblazoned as the stalwart pioneers on AI Mount Rushmore. those who bravely kept up the advance of humanity and art, despite often crushing disappointments, like, death taxes and the lack of sufficient chocolate and the refusal of Loras to operate in convenient fashion)
Currently my favorite checkpoint
Glad you like it!
underrated
Will you make a v2 ? 90% of checkpoints produce anime characters without Lora and I really love this one.
I might bake a VAE into it, but I don't feel a need to change anything on this checkpoint. Most model merges on this site are just chasing user engagement through novelty, and this one does what I want. I focus primarily on Loras as my novelty and I would prefer my results to be consistently applied to one model that doesn't change.
Looks very good. Do you really not plan on doing any more on this? I'd say it's good as it is, but I'm curious.
I might bake a VAE into it, but I don't feel a need to change anything on this checkpoint. Most model merges on this site are just chasing user engagement through novelty, and this one does what I want. I focus primarily on Loras as my novelty and I would prefer my results to be consistently applied to one model that doesn't change.
@Sophorium Thank you very much for your reply, and the model. I myself baked "luna-xlilnai-vae" into it, and I think the results were a bit more appealing. It's basically a higher contrast, I assume. More prominent darks and shadows. That said, base sdxl vae is also fine. A bit brighter overall. Thanks again, and godspeed.
Can you name those (list) checkpoints please ? that you have used for this merge ?
Thank you