I use forge with autismmix confetti checkpoints. I think Jimwalrus has written a good guide somewhere on the forums for noobies with step by step instructions.
Just to build on what I said above. These are images from the game I'm working on. Both have the same loras, same strength, same checkpoint, same prompts, etc. Nothing has changed, but they're not completely consistent.
To people who make AI images, I'm not saying anything new. For people who don't know, these are the kinds of things that hinder your speed. Now, if you're making games and you don't care about quality, it won't matter. I do care, so it's a bottleneck.
Sometime (and feel free to correct me if this has already happened), someone is going to use AI like this to create a game that will utterly blow us away. And will produce it blindingly quickly. Someone will create dynamic scenes with new characters that don't feel like they're cookie-cutter copies of old ones (like every amazon SSR7 draws). Don't get me wrong, I respect the difficulties SSR7 faces in having to draw all of these characters in all of these permutations.
But seeing this, I find myself wishing you could combine whatever AI you're using with SSR7 to revamp the whole of Something Unlimited. I don't know how long it would take, but with AI replacing all of the scenes based on the art that's already there...I suspect it would be shockingly short. Unless there's like 100 bad images for every good one you show us here.
OK, figured I'd just put a link to my post in the thread so that anyone can see it (saves me getting DMs asking for DMs!).
The post linked above is a little old, it was back when a lack of knowledge in Python meant I was stuck with SD1.5 due to some choices I'd made during initial installation.
Anyway, I've now full upgraded to SDXL / Pony (with forays into Flux, but not for generating SR7-like images) and rarely revert to 1.5.
For details how to start using Stable Diffusion, earlier parts of that thread are as good a place to start as any. Although I would recommend using Forge rather than A1111 as the latter is pretty much dead.
OK, figured I'd just put a link to my post in the thread so that anyone can see it (saves me getting DMs asking for DMs!).
The post linked above is a little old, it was back when a lack of knowledge in Python meant I was stuck with SD1.5 due to some choices I'd made during initial installation.
Anyway, I've now full upgraded to SDXL / Pony (with forays into Flux, but not for generating SR7-like images) and rarely revert to 1.5.
For details how to start using Stable Diffusion, earlier parts of that thread are as good a place to start as any. Although I would recommend using Forge rather than A1111 as the latter is pretty much dead.
I'm using the PonyRealism checkpoint and a vixons pony style(detailed v1) lora but I always get some weird aspects like the tiara and necklace which is tough to fix but as far as the face is just inpainting.. I'm pretty new at this but here are the links for the checkpoint and loras I used for her.
bubble butt, perfect ass, spread legs, pussy, sitting on floor, ass focus, laying face up, looking at viewer
you can throw a (spread ass) in but then u gotta deal with wonky hands sometimes or (feet) and (from below) if u want her sitting on feet and showing off bare feet
rest of prompt was just lora for frosty girl and background info
but thank you!! I'm mesmerized by the matching plug
score_9, score_8_up, score_7_up,score_9, score_8_up, score_7_up, 1girl, kfr0st, light blue short hair, blue eyes, blue lipstick, smiling, looking at viewer, dark theme, volumetric lighting, dark dungeon, brown couch, blue buttplug, heart buttplug, <lora:add-detail-xl:0.6> <lora:Anime Summer Night Style SDXL_LoRA_Pony Diffusion V6 XL:0.4> <lora:buttplug_xl:0.4> <lora:KFrostPDXL:0.8> zPDXL3