Made the move to Stable Diffusion after Midjourney nerfed all my best prompts.Seems to have been after the auto-moderation update, but I flat-out can't create messy images on there anymore without triggering the NSFW filters.

First impressions are that it's incredibly powerful, and I'm glad I made the move. SD definitely isn't a WAMmer iby default, and initially I just couldn't persuade it to do anything messy, but in the end I found that controlnets combined with crudely drawn blotches in an image editor worked well. Super glad I moved as it's an order of magnitude more flexible. Being able to swap things out into image editors and then smooth everything over with another low-denoise model pass is super useful.

Another great advantage of having controlnets, LoRAs etc is that you can maintain coherence across images. This is still s WIP for me, but having the ability to create a clean "before" image and then add the mess after is great. My personal kink is super fussy about outfits, so being able to focus on the subject in the initial image and then do a "messy pass", controlling roughly where the mess goes, is a game-changer. Controlnets also add the possibility of photosets and video, though I need to improve the coherence a lot more before that will look any good.

This first batch of images are all very stylised, which I know is not to everyone's taste so apologies if that's a turn-off. It's partly a personal choice, but also because I find the photorealistic checkpoints a bit soulless for the images I want to generate, and even with model swap I found the "realistic" ones less appealing. I'm sure I'll make progress on this as it's only been about a week so far.

For what it's worth, I've been doing a lot of thinking about the stylised/realistic split in ML-based images. I think it'll be interesting to see how people split in terms of their personal "uncanny valley". For me, most of the appeal from this isn't to duplicate what IRL producers can do, but to extend things in directions that are either impossible or impractical IRL. I think there's something truly wonderful about everyone being able to explore previously-unreachable parts of their kink-space.

Great to see you making good progress down your own chosen path!

My own path is exploring realistic wetlook images as you've probably seen, but I've come across a couple of models on my travels that might be of interest to you if you haven't already checked them out:

Amixx
DreamlabsOil_v2

Amixx produces images similar to those in many computer games (i.e. semi-realistic), whilst Dreamlabs mimics oil paintings. I intend to experiment with Dreamlabs myself at some point, but you might like to check it out too.

I can't vouch for how either of them will be with messy content of course, but finding out is half the fun!

Stable diffusion gives so many creative choices for checkpoints and LoRas.

Getting coherence between images is my goal too.

I like seeing stylized images. They are really eye catching. These have nice contrast and pastel colors. Nice RWB colors.

Generating the impractical is my favorite part. No producer wants to spend thousands on outfits, wrecking furniture, staining floors or carpet, wrecking an off site location, or traveling to some exotic location. I'm not into clothing, but I can imagine it would be fun to see what a $3,000 designer dress looks like in WAM scene. Some of the dresses Wetmaxiskirts generates looks quite expensive.