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
Basically this is v5.1 with its DMD-ness heavily neutered. You can use a wide range of samplers, schedulers and CFG if you'd like.
The settings I like are:
Sampler: dpmpp_2m_sde
Scheduler: exponential
Steps: 35
CFG: 5.5
FAQ
Comments (21)
I'll soon be deleting all the other checkpoints from my computer :-) Bro, you're putting out some quality stuff here RESPECT
Great news!
thanks for v5.1 with its DMD-ness version!!!
I will test it soon
https://civitai.com/images/109308858
Could you also test this model as well.. really interested in feedback about the textures with DMD-based models. Maybe you can cross-post the comparison with PROs and CONs?
Please continue making DMD versions
I will always make DMD versions first.
My No 1 model to go MOP
This version 5.1 is close to perfect (with your recommended settings on Fooocus). It notably resolves the smooth texture issue with the Improve impaint on faces. Thank you so much!
Awesome, glad to hear it. I haven't tested this version on Fooocus yet.
@GBRX Everyone says Fooocus is for beginners, but I've tested tons of web UIs and NOTHING delivers such good, and especially predictable, results without any hassle. With your MOP, on a 3080ti 12GB, an image takes a second and a half—not enough time to click Skip while the image is forming if it has flaws. Fooocus's inpaint is very efficient; just one thing: for the IMPROVE inpaint, you need to switch to Karras/LCM instead of Exponential/LCM to get the best results. And it has all the advanced settings and techniques of Fooocus, including its Controlnet and Variations function. MOP works very well with the few personal LoRa "characters" I trained with SDXL on Civitai. Once again, thank you for your fantastic Checkpoint.
@fantastic8 Thanks, I really like Fooocus too. Sometimes I use it just to create a well-styled prompt and then try that in Comfy too.
Most modern SDXL checkpoints usually only generate sub-par garbage IQ and realism with my overly-detailed prompting, but this one is actually quite stunning and super-easy to work with.
Thanks for making a non-DMD2 version. First time I've enjoyed working with SDXL in a very long time!
Great, I'm happy you're liking it
I am wondering if you know anything about this:
Are you able to train character loras with this checkpoint? The reason why I ask is because (so far) I have noticed that, whenever i train characters with checkpoints that have the dmd2 baked into them, those character loras always produce bad fuzzy washed out images (the kind you get when you don't add vae). So, I am wondering if you have ever experimented with training a lora with your checkpoint, or if you are familiar with the issue I have.
Because I am able to train loras on checkpoints that don't have the dmd2 baked into them just fine.
Checkpoints with integrated DMD do not work well when creating Loras. My favorite checkpoints for Lora creation are Big Love and Analog Madness.
@GBRX That's what I thought. Thank you! The two you just mentioned are the only ones I use as well lol.
Appreciate the NON-DMD version, please consider doing a NON-DMD version for future releases. Thanks.
Yes, I definitely will. I'd appreciate it if you could also provide feedback on the Flexible version that I recently posted. I'm trying to provide the textures and behavior of the non-DMD model with the speed of the DMD one.
Thank you for your work! MoP (Mixture of Perverts) DMD is absolutely amazing 🔥
Could you please share which services or nodes you use to create your prompt (NSF/NSFW)?
Great, glad you like it! As far as prompts, they either come from my brain or I borrow them from other images that get posted to civit.
DMD 5.1 is just amazing. Thank you for your work.
So good with prompts, every result is realistic and natural. The way it handles lighting prompts is fantastic. Just the best!
Thanks!