MoP Mixes are a series of checkpoints that merge the MoP (Mixture of Perverts) model with one other model that I find interesting. These are completely separate from the normal MoP releases because those are a much more complicated merge that requires a lot of testing. MoP Mixes are much simpler to produce with the goal of creating a MoP variation that hopefully some people may find useful or just fun.
To be clear, these MoP mixes should in no way be considered better than or a replacement for the checkpoint being merged with MoP. I strongly encourage you to download and test out the checkpoint for yourself and use it in the way that its author intended.
Each checkpoint will be set up in the manner that I think works best. So some of them will have the DMD2 LoRA fully integrated, some partially and some not at all. The recommended settings for each mix will most likely be different so check the About This Version section for all the details, as well as the recommended workflow.
Note: If you want to convert any of these DMD checkpoints to a non-DMD checkpoint then all you need to do is add the DMD LoRA and set the strength to a negative value. I would start with -0.85 and adjust from there. That will allow you to use a higher CFG and different samplers than LCM.
Fooocus Setup Instructions
1) Click the Advanced checkbox.
2) On the Settings tab, leave the Preset as 'initial' and the Performance as 'Speed'.
3) On the Models tab, set both the Base and Refiner models to MoP.
4) Change the Refiner Switch At setting to anywhere from 0.75 to 0.90. I like 0.85.
5) On the Advanced tab, change the Guidance Scale to 1.
6) Change the Positive ADM Guidance Scaler to 1.
7) Change the Sampler to 'lcm'. You can leave the Scheduler as Karras or Exponential.
8) Set the Forced Overwrite of Sampling Step to 7 to 11. I like 9.
Description
This is a merge between MoP v5.1 Flexible and PornMaster V8 which I've never used before, even though it's one of the most popular models. Use the workflow from this image for best results.
This has a little bit of the DMD lora merged into it so it can complete an image in 8 steps. However,, like MoP v5.1 Flexible, it also has some flexibility in the samplers/schedulers and likes the CFG to be around 2.0. The recommended settings are:
Sampler: dpmpp_2m_sde
Scheduler: exponential
Steps: 8-14
CFG: 1.0-2.5
FAQ
Comments (12)
Nice! Pornmaster Pro SDXL v8 is an absolute gem, even for regular SFW stuff. I'll probably have to do the negative DMD thing to enjoy some higher CFG/steps (and to use Euler Ancestral), but really curious how this mix works either way.
Hey, thanks for checking out my model. This one is a bit of a hybrid between DMD and non-DMD. You might want to try it out without negating the DMD at first, testing different samplers, CFG levels and steps. It should be fairly flexible but if you need to negate the DMD then I probably wouldn't go any higher (or lower?) than -0.5 on the DMD lora strength.
@GBRX I'll def give it a try normally with DPM/Euler, but unfortunately my rig's too old to get any software with LCM working, which is ideal for accelerated models imo. Just had some pretty good success w/enzino's IntoRealism Ultra "light-DMD" v11 with Euler and really low CFG/steps and also with neg-DMD and normal CFG/steps (-0.4 to -0.5) - using negative DMD actually looked better, which was weird.
Usually, the biggest problem I run into is when I'm trying to use a character LoRA with low CFG on an accelerated/turbo model and the results bare little resemblance even at max strength and using emphasis on its trigger words. IntoRealism actually wasn't too bad in this regard, but it was still an issue.
@GBRX Just to update, you're 100% right! Both with and without negative DMD LoRA are both viable options with this model. Using Euler A, I could get comparable results using -0.25 to -0.40 with high CFG/steps, but it's much trickier to dial-in for any given prompt (or LoRA load-out) and it's just easier to use it normally (still w/Euler) at your recommended CFG/steps. Once dialed-in w/neg-DMD, I can sometimes achieve better fidelity and adherence than without, but imo it's just too tricky to recommend unless someone just likes experimenting.
And your new MoP Pro v2.1 can also work with DPM/Euler and high steps w/neg-DMD, but wow, that's super difficult to dial-in just right and think you were right to only recommend LCM for that branch. I'd wanna "port-beg" for a non-DMD version of Pro, but I'm guessing the quality/realism boost it's getting by using LCM/DMD at low steps would mostly be lost in a non-DMD version (though I'd love to be proven wrong!) - anyway, nice work on both!
(edit, forgot to add, this Mix works very well with LoRA adherence in either mode of operation, so kudos there, too!)
what is the node named 628caf29-87b2-45be-a3bc-3c864e230cf8 used in your pornmaster mix for use in comfyui
Looks like that's the subgraph that handles the upscaling
Auch wieder Dreck!
More specific feedback would be helpful
@GBRX Schau in mein Profil rein da ist ein Bild das ich eigentlich hier hochgeladen hab aber nicht angezeigt wird.
@teeeta43835 Sorry, I have no idea what you're talking about. Sounds like a CivitAI glitch. You should open a support ticket.
Guck dir mal deine Bilder an und dann denk noch mal drüber nach wer hier Dreck verzapft :-) Übung macht den Meister.
Omnia is still probably one of my favorite checkpoints ever.
Any plans for a Omnia V2? It would be great to see it native 1536px support or something like that :)
Nonetheless, thanks for your great work!!