This was something I made mostly just to see what would happen. (That's why I made the Powerpuff Girls and EQlvl1 loras)
What I wanted was a realistic model that I could use without always loading in my MBH lora, and why not a few other things while I was at it.
The first 9 images on all versions except AnimeConversion are the same prompts, seed, samplers, CFG, and steps, etc.
These are three unique models!:
MBHRealityPlus2 is an update to the original focusing on SFW and NSFW renders.
MBHSemiReal puts a slightly realistic twist to an animated look.
MBHAnimeConversion pushes the animated look nearly all the way to realism even with anime LoRA's.
The connection between them is the use of MBH-Pony in the merges that created them.
Read "About this version" on each for more info.
Description
mklanPony - Juggernaut merged with add detail (1.0 strength), MBHPony (1.25), age slider (1.0 because I wanted the ages to look closer to the specified age... also only about 75% certain that's the one I used), a pussy lora (0.20), and a dynamic pose lora (0.20)
This model works well as a solo photography model for clothed and nudity but struggles with truly NSFW rendering because of the 2 NSFW loras baked in focusing on specific body parts and types. I'm sure there are plenty of "position" loras you could add to make it easier if that's what you're wanting.
images rendered in ComfyUI using pony prompting (score_9 etc...)
Suggested CFG 2-3
steps 20-30
But play around with it.
Post your renders please. It may seem trivial to some but sometimes it helps to show creators the hours spent training are worth it.
FAQ
Comments (4)
Looks awesome! But why no prompt metadata in the images? 🥲
I'm going to be honest and say that if that's something that can be done on here with images created in ComfyUI, I'm not sure. In theory though you could download them and drop them into Comfy. I jumped onto the bandwagon quite late so I'm still learning it myself.
edit @Icey64 There's one image now with proper metadata and unless there's a node I don't have / can't find, there's no metadata on the others because you have to input the data manually at the time of saving. It's quite tedious to do it on every generation and does not account for upscaling and FaceDetailer nodes.
Also I see now the other's have been converted to JPEG's, that's not what was uploaded, but at least there's one with the data.
@axicec That's actually what I have/used. I just had not thought about converting widget to input. Partially because now my nice neat (minimal nodes) workflow is messy and convoluted. I'll look for a way to streamline it though and maybe look into switching from "Tiny" nodes to the pipes series, perhaps they will work better with that image saver. But I really like my tiny nodes and the clean straightforward flow LOL.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.