This is a checkpoint merge of multiple models with the specific intention to keep features that translate well from VRC trained loRAs to semirealistic/anime pictures.
It works pretty well turning 3models similar to VRC into anime and vice versa. Results outside these fields may vary. No extra VAE required, any amount of steps over 20 should be fine.
Recommended negative prompt : bad quality,worst quality, worst detail, sketch, monochrome, lowres, text, error, cropped, low quality, jpeg artifacts, 3d, cgi
Base models included in merge are all merge allowed.
Description
FAQ
Comments (6)
I'll be honest - I wasn't expecting much from this model just by looking at your example images.
However, it surprised me by being pretty creative while still adhering well to the prompt. The generated images are interesting!
You made some really cool stuff with it! Thanks for giving it a shot.
I can't see any of the cool stuff you made with this model in gallery anymore?
@Vencenzo Someone went through many of my images and reported quite a few. Rather than fight, I decided the number of images I had were far too many anyway so I deleted all older than 21 days with less than 20 reactions, or older than 10 with less than 5, or older than 3 days with less than 2.
My gallery count dropped ~98.5% which means either the majority were trash anyway, or most images never really got more than 1 or two people to look at it.
It also means that many models with small number of images, that had a large share of mine, are now emptier.
Long (and boring) story short, I’ll gen some for yours again.
@clueless_engineer Wow, that's really sad. To avoid situations like this, I download my favorites from your generations locally to my NAS. That way, I can go back to them in the future.
@clueless_engineer Thanks and sorry to hear that. I did at least download 2 of my favorites of yours from before. I have a base model I am working on that I have been baking in all the enhancement loRAs I use. Been messing with tweaking weights so it looks just like I used the loRAs with only the base. Will post that soon.