Photorealistic AI Generated Images

[SUNBO]

Thanks for the info, I got it setup and have tested out a few different models. Seems like most of them are for making anime, which is not really my thing, personally. Through some reading about this I see that a lot of the nsfw content has been marginalized on civitai, but as you say, it's not that simple. Seems like it just takes a lot of trial and error. Are there any models in particular which you have found good? I'll try that nikon z9 prompt, I see a lot of people using "masterpiece" online.

You need to sign up to civitai to see the nsfw models. There are alot of realistic models on civitai. I use qgo10b for model alot.

 

[zombierambo]

You need to sign up to civitai to see the nsfw models. There are alot of realistic models on civitai. I use qgo10b for model alot.

Thanks, I see it here, I'll give that a shot. https://civitai.com/models/4188/qgo-10a

 

[maload]

i think i can do it with model and googlecolab .
there is something in my mind
this ai can change their face or not ??
which factors that create their face ??

 

[SUNBO]

i think i can do it with model and googlecolab .
there is something in my mind
this ai can change their face or not ??
which factors that create their face ??

You can use whatever face you want. But I haven't looked into how, since I'm not interested. You can watch some youtube to learn. https://www.youtube.com/results?search_query=stable+diffusion+face

 

[maload]

You can use whatever face you want. But I haven't looked into how, since I'm not interested. You can watch some youtube to learn. https://www.youtube.com/results?search_query=stable+diffusion+face

thank you so much

 

[intrepid8]

Be aware that you can mix models together in SD/A1111 and give them weights, so if you like some aspects of a specific anime model (A) but generally prefer a realistic look that you get from model (B), you can merge them to form a custom model like B (75%) + A (25%).

 

[zombierambo]

Some new generations, after downloading some new models from civitai

 

[Zen10101]

I'm giving my GPU a rest after seeing what the hi end pros are capable of



Please have them scan prime Riko Tachibana and Yuma Asami to continue their legacy...then movie stars, holy hell. We're only a few months in, next few years is gonna be wild.

 

[Casshern2]

I'm still very new to this and I'm absolutely relying on online resource, I don't have a box strong enough for Diffusion, but one thing I notice on this thread (slightly) and on so many others out there on the web is that most things that are posted have basically the same look. The face is the same, the relative shape of the model is the same as is basically the same pose/position, just in different outfits and varying degrees of breast size and shape. I'm all for having a favorite actress in JAV and can watch some do anything in any setting in any outfit, so maybe that's it? Whoever is posting out there has a favorite model.

But variety is the spice of life! I had started a thread (here) which I won't keep up since this one has more traction, don't need two threads for the same topic, but even on that one the examples given by @Zen10101 especially the video speak to what I'm saying.

One online resource I found is really great at creating prompt-driven (as in buttons) images and I've been able to get some amazing "real-world" looking results. The native resolution for the generations, though is 512x512. They do include decent upscale feature because of this. I sometimes use it or I'll try others. The results are good enough for me, maybe not others. And the models that are generated, again...do it for me. Of course, there are still glitches and hands/feet/toes don't always look appetizing. Plus I can still find artifacts after trying to fix/change things. But since I'm after arousing images, it's fine.

Native




Upscaled

 

[Casshern2]

Here are some recents

 

[Casshern2]

Well, even this platform I've been using "suffers" (air quotes intentional) from similar looking output, so that means most if not all its generations are model based, pretty much? Still, though, great output!

 

[IrotsuyaSan]

I prefer older, but this is exception

 

[IrotsuyaSan]

"She isn't real. I can't make her real"
Slipknot - Vermilion

 

[Zen10101]

 

[Casshern2]

 

[Casshern2]

I certainly hope others are trying this, you can get some amazing results. "Photorealistic"? I guess depending on what platform you use, but at times, no, as evident in my examples, you can always spot flaws. But for all us used to letting your imagination fill in the gaps, good grief, the results!

 

[Casshern2]

I'm telling you, some of the women that get generated are insane! imo

 

[barba]

Some new generations, after downloading some new models from civitai...

just the thing for this weekend at the indy 500.

 

[Casshern2]

Depending on the platform you have plenty of prompts to choose from, like age, hair color/length, type of clothes or accessories. Setting, like the street or a house our a classroom to name a few. All kinds of poses and possible camera angles. It can be overwhelming at times thinking what you want to see, and once you see it...what more can you see!

 

[Casshern2]

You can generate skinny women, built women (toned, athletic), or thick or plump, big hips, boobs. Mature, young JAV debut ages, whatever you feel like seeing. You can have them wear a lot or a little or nothing at all. Place them in classrooms, outdoor bubble parties, the street, in all kinds of positions and postures. Something for everyone.