DPM++ 2M SDE Karras
DPM++ 2S a Karras
DPM++ SDE Karras
DPM++ 2M SDE Exponential
DPM2 a
DPM++ 2S a
DPM++ 3M *
Euler A
Steps: 30 or more
CFG Scale: 6–7
Clip Skip: 2
Resolution: Greater than 1024px
Hello, I hope you like my model. From here, it is a privacy for my realistic models. Mix it all together don't want you to lower my points . So I have many of my models that look realistic... mix them here.
Please, comment, I hope you like them... AD
Description
FAQ
Comments (5)
Great model, but you posted the same model as version 1 and 2. please post the correct model so we can try both out. These generate identical images with identical prompts.
@MagicArt35 Impressive work with vII, This is hands-down your best work to date (at least out of all your realistic models that I've tested), so absolutely another step in the right direction and a model I'll be revisiting again. Everything seems improved over v1 and LoRA adherence is one of the best I've ever seen, especially with character likeness. Think anime-based LoRAs still have a slight effect on image tone/color/style, but that happens with most photo realistic models and also seems improved over v1.
One weird thing I've noticed between both models (and a few of your prior works) is while it does an excellent job with prompt adherence at common CFG values, it has a strong tendency to completely ignore negative prompting even when weighted. Not needing to use negatives is a nice ideal and I'm thinking that maybe this is an intentional design decision, but sometimes that's just the easiest way to easily remove an unwanted occurrence of something without unnecessarily complex positive prompting. And sometimes it almost seems like it's interpreting some of my negatives as positives, which is kinda strange as well.
Overall, it's a very fun model to work with. Thanks for sharing!
@MagicArt35 Just to update about the random crashing with CybeNatuXL v1 and vII, I still haven't pinpointed exactly what is causing it, but I noticed it only happens while loading a prompt. No error messages are shown in the console, just a generic pop-up saying "python.exe has stopped working". I can then restart my software and use the exact same prompt with a different seed and have it work just fine. Very strange! Doesn't happen as much with v1, but with a full day of testing, it became quite regular with vII. Was thinking it was maybe the larger size of vII causing a memory issue, but that wouldn't explain why it occasionally happens with v1 as well (everything on this machine has been very stable lately, so don't think it's a hardware issue on my end - @Tarendax Did you have any crashes with the 12GB vII of this model, or just me? Thanks!)
Not really sure if that's helpful info or not, just sharing my experience. And while it's pretty annoying to deal with, it's a small price to pay for how good both models work otherwise! :)
@AFD_0 That hasn't happened to me;
@Tarendax Thanks! Must be something weird on my end that it doesn't agree with.