This is an Illustrious version of my photorealistic MoP (Mixture of Perverts) DMD checkpoint. Basically it is just MoP v4.0 merged with some Illustrious checkpoints and LoRAs.
With v1.0 I tried to balance between getting a good semi-realistic look while still keeping some of the lighting, color and composition benefits of MoP.
v2.0 is intentionally much more cartoonish.
This is a DMD model, intended to produce images very rapidly. I recommend using my SDXL DMD workflow.
The recommended settings are:
Sampler: LCM
Scheduler: karras, exponential or beta
Steps: 6-14
CFG: 1.0-1.3
Clip Skip: 1-5
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 an Illustrious version of my photorealistic MoP (Mixture of Perverts) DMD checkpoint. Basically it is just MoP v4.0 merged with some Illustrious checkpoints and LoRAs. I tried to balance between getting a good semi-realistic look while still keeping some of the lighting, color and composition benefits of MoP.
FAQ
Comments (27)
lets go!!!!
I love your checkpoints my friend!! This one is GREAT!! Thank you! Well done!
Thanks!
Great work! Keep going on! Congrats.
I'm a big fan of the regular MoP merges, I've been using 10alt and 31 a lot. That said, this particular checkpoint gives me very plastic-looking faces. Maybe there's some trick to prompting Illustrious merges that I'm not familiar with, or a LoRA I should be using, but so far I'm not a fan of this particular merge.
Hey, thanks for the feedback. This model is just kind of a fun toy that I made. It's supposed to be semi-realistic and a little cartoonish while still retaining some of the color and lighting qualities of the SDXL MoP model. If you want true photorealism then the SDXL version is the way to go.
@GBRX Awesome, thanks for the reply! I see a lot of people using pony/illustrious merges and getting that semi-real look, so I realize that must be what some folks are shooting for, but it's not really my thing. Anyway, I appreciate your experiments, you've hit on some really nice stuff so don't stop what you're doing!
@kgamer1245 Thanks again. I also much prefer photorealism but semi-realism can sometimes be fun too.
@GBRX would you recommend any realistic illustrious checkpoint? does something like that even exist? Illustrious is "cartoony" by default but i like it.
@ElectricDreams Yeah, I wouldn't consider any of the Illustrious models to be truly photorealistic in the same way that a good SDXL model can be. The point to Illustrious is that it can be more colorful and visually creative than a SDXL model. I linked Perfection Cinematic in the suggested resources for this model. I like that one for what it is.
@GBRX oh i like your checkpoint a lot and i hope you keep going on with illustrious, make no mistake. I don't mind the cartoony looking, i was just saying it because i read the comment of "very plastic looking face" when that pretty much comes with the territory. So i don't think is a very valid criticism, tbh.
@ElectricDreams Nah, that was good criticism. I probably could have been a bit more descriptive about what kind of images this model is supposed to output.
@ElectricDreams hi,
try FaserCore :
https://civitai.com/models/679856/fasercore?modelVersionId=1048561
here one of my gallery :
I really liked it, let's see if I can take it to the limit, great work.
I really like this one, I think is better than SDXL version and the DOF is not to much.
Good job!
Thanks!
yeah, this is a really good model. It took me a bit to get what I wanted out of it but once I figured out all its nobs it is one of the better realistic models.
Good results, but TOO juicy for my taste) Excessive contrast, striking detail in the eyes... too striking, sometimes it seems that the eyes want to become anime. Oh yeah, 10 steps and 0.7 CFG was the best setting for me.
Thanks very much for the feedback. I definitely agree that this has some issues. I went a bit overboard. Although I do like how many of the images come out, it's a bit unstable of a model. I'm working on a version that I think will fix many of these issues.
@GBRX Wow, I'll be really looking forward to it)
@GBRX I ADORE this model!! I sprinkle it on all my Checkpoint "food" and it just adds that Umami flavor that I'm looking for... and, I too am excitedly awaiting an update of V1!! - Geno
This model is very realistic, but you quickly reach the limits in terms of possibility.
If I want to have correct images I have to do image to image, otherwise the renderings are always the same, same position and why always by the pool?
I think is biaised...
Maybe the Guidance Scale 1 is too low, 4 or 5 would be better for a new version.
I would prefer a version that is a little slower but that respect better the prompt...
Thanks so much for the feedback. It's a DMD model so a higher CFG won't work.
It's my first attempt at an Illustrious model so I'm sure it has lots of room for improvement. I'm not sure what you mean by "always by the pool" though.
@GBRX what's a DMD model ?
@GBRX I mean by 'always by the pool' is that if I ask in the prompt for example 'wet skin' he makes me an image outside near the water (pool, beach...) I think he doesn't understand the words I use.... this for the 2 versions of MoP
Do you have the possibility to improve this?
@Suzanne DMD2 is a highly trained lora that adds a lot of realistic detail but also speeds up generation. https://civitai.com/models/1608870?modelVersionId=1820705
@Suzanne I see. Are you specifying a location? If you just say 'woman, wet skin' then it might use a pool as the location but if you say 'woman, wet skin, shower' then it shouldn't use a pool.
I don't really know if there's anything I can do about that behavior though.