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
Honestly, I can't say that this is significantly better than v4.0. I've actually reduced the DMD2 strength about 20% which I think helps reduce oversaturation and makes image generations a bit more consistent. I may reduce it further with the next version.
I toyed with just calling this v4.0 Alt, labeling it just an alternative version. However, there are a lot of changes so I think it warrants being called v5.0. I can probably still squeeze a little more quality and consistent photorealism out of SDXL but probably not much.
The recommended settings are:
Sampler: LCM
Scheduler: karras, exponential or beta
Steps: 5-12
CFG: 0.8-1.3
Clip Skip: 1-5
My favorite setting is LCM\Karras, CFG 0.8, 7-9 steps. It can be a little unstable and I might need to generate more images to get a really good one but I find those settings can create great images.
FAQ
Comments (15)
I still haven't used this checkpoint but if you're having problems with color the best thing is to use (merge) these loras for total color control (doesn't degrade IQ) https://civitai.com/models/1363766/sdxl-color-tweaker-slider-color-temperature-saturation-brightness-contrast-or-sdxl
Incredibly realistic images are presented! Very impressive!
You should merge with some saturation and color lora to balance a natural photolook. BigLove become slowly a cgi monster.
I use this style of photorealism for a reason, to make start images for WAN I2V videos. I find this style works really well for that purpose but I'd be happy to take a look at any specific Loras you think would improve it.
@kyreartyom823 The Checkpoint ist Great. Need more color? Install: : https://github.com/hahnec/color-matcher/
@kyreartyom823 FYI, I just published GonzaLomo v6.0. I think that will have much more of the balanced photorealism you're looking for: https://civitai.com/models/1513492?modelVersionId=2368123. I'd be happy to get your feedback on that if you'd like to try it out.
Also, I'll be applying a similar process to future versions of MoP to make a bit more balanced as well but it still needs to primarily work well for video start images.
Again, great work. Less colorful, thanks for that -)
So we don't have to use the DMD2 LoRA with this one? It's already included? Thank you
That's right, DMD2 is already integrated into the checkpoint. Just be sure to use the recommended settings.
@GBRX Thank you sir, I'll give it a try!
v5 can do some stuff, but it's a bit worse than 4.0 on the saturation, and can get a bit plasticy or have excess detail in features. It also seems to have an issue of wanting to give dark spots on skin and even spot like artifacts in textured things like cloth. It just seems baked wrong, and removes a lot of the realism. Under certain circumstances it is better, but it has issues.
I think the DMD2 strength change could be part of the culprit, because it's probably in some between area where the sweet spot is in between steps, so you can either choose 4 steps or 5 steps, but the sweet spot might be 4.5 steps, except you can't do 4.5 steps.
Thanks, I'm already working on a v5.1 to address all the issues.
@plk I released a v5.1 version that I think addresses these issues, https://civitai.com/models/1854124/mop-mixture-of-perverts-dmd?modelVersionId=2377298
@GBRX Maybe you can take a Contrast and Sharpness modification LORAs to make CFG range wider. The contrasts seems to high.