This started as a purely experimental merge, separate from my GonzaLomo models, but has now become my main SDXL source of images. I think the name Mixture of Perverts says it all.
My primary purpose in creating this model is to generate images that I can use for WAN I2V. I wanted something that had a more amateur feel while still being able to produce good color, composition and lighting.
The v5.1 models are among my favorites and offer a wide variety in speed and usage while giving the best photorealistic results.. See the version details for the recommended settings for each of those.
v6.1 and v7.1 are closely related and offer a decent variety of styles and facial types for DMD models.
Most of these versions are DMD models, intended to produce images very rapidly. I recommend using my SDXL DMD workflow or you can choose to use the workflow embedded in the showcase images.
The recommended settings for the DMD models are:
Sampler: LCM
Scheduler: karras, exponential or beta
Steps: 6-14
CFG: 1.0-1.3
Clip Skip: 1-5 (1 doesn't work with some of the newer models)
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
I think this quick update should address most of the v5.0 issues with oversaturation. This version should also be able to work well with a wider variety of settings, especially CFG. I've gone as high as 2.0 if I set the steps to 20. This version should be able to handle much higher values for both CFG and steps than previous versions.
I welcome any and all feedback so that I can make this better!
The recommended settings are:
Sampler: LCM
Scheduler: karras, exponential or beta
Steps: 5-20
CFG: 0.7-1.5
Clip Skip: 1-5
A good place to start with the settings is LCM/Karras, 11 steps, CFG 1.0, Clip Skip 2.
FAQ
Comments (11)
They just keep getting better. I use these with Fooocus and the results are excellent quality except that I cannot figure out how to avoid harsh light and hard shadows with interior scenes. They all look like flash photography even though I put that (and other terms) in the negative promt.
Very nice! This version is very realistic and not quite so "hot" color and exposure wise. I love your work, well done!
At the moment is the most powerful XL model for me. :fire:
This and GonzaLomo would have been amazing models, but they are both ruined by baked in DMD. I don't need the speedup. I would very much prefer having access to CFG and different scheduler/sampler combos. Those who do need the speedup can easily add the DMD LORA themselves.
Actually, the MoP v4 can bring nice wood texture and sometimes the skin texture, but composition is less complex and narrative not so rich. So, it is good as refinement if you can lower the contrast / CFG by Details Daemon node for comfy or Contrast Slider LORA.
You can use the CKPT of your choice in Text-to-Image mode + the MoP for HiRes Fix or ADetailer pass.
Thanks for the feedback but I disagree about these models being ruined by DMD. Just the opposite, I find the photorealistic images from DMD-enabled models to be much better than those without DMD. I do offer a non-DMD Gonzalomo v4.0 though so you can try that.
Also, you can easily remove DMD by adding the LoRA with roughly a -1.0 strength. You might need to tinker with that value a bit but that will allow you to use different samplers and higher CFG.
@GBRX Not ruined, but textures and composition suffer. I'm trying to bring back the creativity and textures into DMD based models, but still not quite there.
Please pay attention to textures of the ground, fur, skin and grass. Sometimes they appear over-baked and too contrast. Realistic photographs have noise in textures and smaller details.
We need to address it and make it possible to render noizy natural textures. Realism requires noise in textures. Also, wetness seems a problem to me.
this is amazing, i do not have the plastic skin effect anymore, beautiful models
MoP 5.1 is very sensitive but great at inpainting skin and textures on cartoon generations. I started the images I made below with NoobIL checkpoint and put skin texture to them. I have used so many different xl checkpoints to do this and this was the best one so far. I haven't tried it with increased breast and belly sizes yet, but I'm hopeful.
The belly and breasts work well with a lora. Looking like this is the new champ when it comes to switching from cartoon and illustrations.
@BBBenthusiast Thanks! I really like model switching from cartoon to photoreal. It might take a few tries to get it right but you end up with some very creative results.