It's Berry's mix with F222 instead of F111 Edit: At least i guess so, same models included. Edit 2: Just tested it to be sure, seems i am correct: Result "BerryMix14.ckpt [2e41a25c]". Worth it to note that F222 is actually based on SD 1.5, so subtracting that in the "NAI + (F111 - SD1.4)" step instead of SD 1.4 would probably lead to better (but different) results.
It's Berry's mix with F222 instead of F111 Edit: At least i guess so, same models included. Edit 2: Just tested it to be sure, seems i am correct: Result "BerryMix14.ckpt [2e41a25c]". Worth it to note that F222 is actually based on SD 1.5, so subtracting that in the "NAI + (F111 - SD1.4)" step instead of SD 1.4 would probably lead to better (but different) results.
This. I think there was a point where I fucked up and made one of the steps with the clipskip 2. Don't know if it makes any difference, but since it works fine, I didn't bothered to fix.
That's it - Just a Berrymix with SD 1.4 and F222. Follow the steps normally, just change the F111 for the F222.
I'm having a hard time recreating this prompt, do you generate at 1024x1920 or upscale there? Any other ideas for settings or variables I might be missing?
I'm having a hard time recreating this prompt, do you generate at 1024x1920 or upscale there? Any other ideas for settings or variables I might be missing?
I'm generating at 512x960 and then upscaling. I used CLIP skip: 1, no hypernetwork and latent scaling. Also check if you don't have "Use old emphasis implementation" selected. Nothing else comes to mind. I checked the prompt just in case and it gives me the exact same image.
I'm generating at 512x960 and then upscaling. I used CLIP skip: 1, no hypernetwork and latent scaling. Also check if you don't have "Use old emphasis implementation" selected. Nothing else comes to mind. I checked the prompt just in case and it gives me the exact same image.
where and how is this Ai, if it is. bro teach me. I do not think this is AI
It's using Berry's Mix. Just match the prompts and see it appear before your very eyes. I think I did some inpainting to change something (can't remember), but follow online guides and use the prompt and you'll get something nearly identical.
It's funny that I put 2girls in the prompt, but it made 1 girl and a small portion of what I'm suspecting to be a guy. Still... having more or less people than desired seems to be very common.
Huh. My guess would be that it has something to do with how image samples are loaded. Maybe certain colors appear differently when scaled down, but when shown in original resolution, it just appears normally or something.
