There is still much work to be done, but am sometimes getting amazing results.


Blue Boilersuit Girl, Gunged.
DM1, pixels on a screen, 2023.

Good start.

Here's a selection of boiler suit AI generations from a gunge trained model in stable diffusion.

The suit turned out quite flexible in terms of how tight or baggy or whether any cleavage. It's not my thing so I wasn't sure the styling of the boiler suit (personally I'd go for low cut and cleavagey but that may miss the point )

Damn I love how natural the slime looks.

I feel
Like we're getting closer to the "singularity " in wam where we don't be able to tell the difference.

Oh man I wished I saved more "vintage" wam from before AI haha

Those are *extremely* good. The suits look a bit plasticly, more like rainwear - traditional boilersuits are cotton or polycotton - but given the quality of the results that's a fairly minor issue. And the gungings are uttely brilliant, really, really high quality work.

The style and fit of the suits is very good too, and yes, low cut / cleavage would entirely miss the point. The trick with putting girls in boilersuits or jumpsuits is they need to fit well and show off the figure, while being fully covered and practical - and yours do that beautifully. I particularly like the fit and shaope you've got around the waist, hips, crotch, and upper legs, that's the exact fit I look for in jeans, and jumpsuits, on real models when shooting. To me (and I think many others) the sexiest part of a clothed woman's body is basiclly the area from just below the crotch to the natural waistline - something that's really enhanced and drawn attention to by high waisted jeans, shorts, trousers, or jumpsuits, with a long fly. This is why woman in high-waist Levis 501s always look so amazing.

So yes, very, very impressive work.

Our of curiosity, will you be licensing your wam-trained SD model at all?


We're not quite there yet, mainly because so far, you can't create a sequence - you can't create a beautiful person in your perfect outfit, and then have a system generate a series of images of them getting messy - walking down into a swimming pool filled with custard, or being pied and the aftermath, or wading into mud and rolling around in it. And likewise you can't yet generate video of an event like that.

I do firmly believe it's coming though. As I've said before, I see it as the dawn of the age of steam locomotives. The first successful machines were huge and cumbersome and complicated, went very slowly, and initially weren't any threat to the centuries-old and well established horse and carriage business. But they worked constantly and never got tired. And once the idea was proven, more and more new engineers and builders got involved, and within a few years the entire world had changed forever as trains capable of 60mph and more connected the country together, and the horse carriage business was consigned to history.

I don't think AI WAM will entirely replace human modelled WAM, but I do think it's going to make a massive impact, and change the business side of things, probably in ways we can't even fully imagine yet.

Meanwhile, here are a couple more images I created using the same new platform as the one at the top of this thread.


Gunge bathing babes of 2023
DM1, pixels on a screen, 2023


Splosh!
DM1, pixels on a screen, 2023

It's a new AI platform I've had advanced access to. Still in development, but when they hit, they really hit. And this is without any wam-specific model training.


Thank you!

The images in this thread are stunning Yeah, the singularity is definitely approaching! Realistic-looking movies within a year, tops.

This whole tread has got me thinking,

Could we as a community create our own prompts? Like have a database somewhere what anyone can download. Stuff that we know works, to generate these types of images ?

Open source and updated whenever ? Maybe even a pdf of how to get started

Thanks. With my stable diffusion models I'm not against the idea of them being shared. Currently though I trained it solely on my own shot WAM images, so the output is very tailored to the style of those original photos. I'd like to see models trained more broadly on WAM images but I think before that happens we likely need a conversation around how producers feel about images they upload to UMD being used for AI training purposes. If we have the permission of those that shot the images then some good general gunge / pie / mud models could be trained and shared with all, maybe through civitai. My current models produce images very much in my style, so if others used it then their images would look pretty similar and we wouldn't get much variety in the AI Wam produced. Even a poll of producers might be a good start to gauge the feeling around this.