Holli Would
Hey all! This is a character I've thought of doing for a while and I figured I'd give her a shot myself. I actually watched the movie and went frame by frame in parts to grab as much of a dataset as possible and ended up with over 200 images! She's very diverse with facial expressions and poses that I can see with testing. Let me know what you all think! She was fun!
Trained with NoobAI-vPred but works perfectly fine with illustrious checkpoints
Grabbed some dancing and I Found it easier to pull those up by prompting (Dancing:1.2) if you're interested
Some nice facial expressions came up so she's pretty expressive with: Angry, annoyed, and naughty face
Let me know what you all think and post images! NSFW welcome :D
Description
FAQ
Comments (17)
not gonna lie, but had more authentic and better results on realistic checkpoints than on anime checkpoints.
Yeah? I'm curious to see what you came up with though! I'm sure they look good just may not be exactly like the original look. I had to take screenshots of the actual film and my god, the lines are not crisp so I think that made it difficult in certain aspects. But I'm definitely curious to try and see what you have as well. thanks for the feedback!
I actually just tested it with Yetanotheranimemix and seemed to work fine! I'll post it and keep in mind I didn't bother inpainting I'm tired and it's late lol
About half the photorealistic checkpoints want to make her look like Margot Robbie. URPM gave the most interesting results actually, kinda halfway inbetween cartoon and real, but URPM had no idea what to do with backgrounds, those are just a random disaster.
OHHH That makes sense it may not know what to do with those wacky backgrounds from the film lol in Versioin 2 when I end up updating her I'll prompt my own backgrounds in so maybe you can have a bit more freedom :)
Wow I forgot about Cool World.
Haha right ?! What's funny is I have a lonette scheduled to post tomorrow morning and literally just yesterday someone posted one ! I feel like such a jerk 😆
@DeskGrenade Never hurts to make more than one and then compare yours to theirs! it's fun to see what someone did different during training so you can correct things if yours doesn't turn out as well. You can also try another person's attempt and think "whoa. That sucked a LOT" and want to go fix it yourself :) That's why there are so many different versions of the same character on this site :)
What else are you using besides easynegative, verybadimagenegative_v1.3 and Babes20 ckpt? I'm putting these same exact prompts with the negs downloaded and not getting those results at all. Same as the info tells me - steps set to 20. CFG Scale 4. I tried with Comic Babes ckpt too and can't get a cartoon result. Are some positive prompts textual inversion embeddings?
No textual inversions at all actually ! Are you using the babes 2.0? Thata what I use. Are you using high rez fix ? When I use that , I set it to high rez steps of 20 and .50 denoising strength I believe. We can use the chat here and I can help if you want 😁
@DeskGrenade Thanks for the offer but I doubt we could live chat because I work evenings+sleep days. Yes Babes 2.0. I wasn't using highres.fix before but I am now with those recommended numbers and it's not cartoony. I'll upload my highres.fix results. My old Stable Diffusion's Python bugged out and I was forced to reinstall but was met with a strange new format. Now the bat startup says SD 1.7 - There's no option for on this site. Some Loras have worked. Probably going to be hit and miss with Loras tied to SD 1.5 and I can't use SD 2.0 up.
@MoeSSyzlak oh man that sucks ! But I'm glad it's helping using the settings I gave you and look forward to seeing the images 😄
@DeskGrenade Thanks for your help. I deleted my first contribution of images and the last ones are those with highres.fix, 10 based on the same settings all around as your previews. Because of that, I believe the problem is on my end and not this Lora. I guess I'm stuck with forced realism in SD 1.7
@MoeSSyzlak is your clip skip at 2?
@DeskGrenade Yes. When I set it to 1, the realism gets a bit more extreme. Like a photo. Can't seem to get a cartoon on 1.7 except with anime ckpts - and those have the typical anime eyes.
@MoeSSyzlak that's so weird that 1.7 does that ! I know its a pain to install but can you go back another version?
@DeskGrenade Nope. The old version was throwing up too many errors. Wouldn't load. It was unusable. Thanks for your help though. I learned a couple new things.