This model was merged by mixing what I considered the most realistic models together to create an even better base model.
VAE is built In
Creates anything from Portraits, Landscapes, to NSFW
Keywords
Imperfect - adds skin imperfections like moles, scars, and acne
Nude - should produce a model with no clothes
Architecture Photography - Used to create buildings or structures
Wildlife Photography - used for animals and nature
Car Photography - used for roads and vehicles
Food Photography - used for plating and food
Interior Photography - used for interior structures on buildings
Landscape Photography - used for portraits of nature
Hyperdetailed Photography - When you really want details
Cinematic - Make it look like it belongs on the big screen
Full Body Photo - When you want the whole body in the shot
NSFW - Just use typical NSFW Jargon and it should produce what you want
Realistic - Does exactly what it says produces realism
I use any DPM model and I use a CFG of 5 with steps around 25. The lower the CFG the more realistic. The higher the CFG the more creative AI becomes. As you increase CFG if you start to see artifacts increase the Steps.
This is my first time uploading so feedback would be appreciated.
A big thanks to the creators of Juggernaut, Imperfect Portrait, Pyros NSFW, and Realistic Vision as those are the models I used to merge with.
Description
Minor adjustments added some pony to it for creativity so pony models should now work with this model. Also added some more realism to the model.
FAQ
Comments (9)
is the LCM LoRa still built into v4 or was v4 built from v2?
I added on to version 3 so the lcm is diminished slightly from v2
One of the better models...
Thank you for the kind words....working on big things soon with it. I am going to start toying with 3.5 and Flux but in my opinion SDXL is still king of realism for now.
@terrariumbuilder497 great to hear -- even in my simple attempts, it's not so straightforward to make merges or trains as well as your series has been. I've been moving a lot of SDXL models, and yours has survived for a long time. Even as bigASP and these Illustrious models emerge.
@makiaevelio543 Time is a big killer for sure with 3.5 and Flux. Also alot of people don't have the means to train them so that is holding those back as well.
@terrariumbuilder497 yeah most of our consumer cards get bumped out. It's hard to consider even trying to improve Flux when you can barely output a picture lol
Any Flux plans? (just saw the other comment)... I'd love a Flux version of this, even though it's got issues.
Working on flux and sd3 versions but my PC isn't quite where it needs to be








