These Women Do Not Exist ..... The NSFW A.I. Art Open Thread
- Thread starter Duece
- Start date
Artificial intelligence
Lol, where did you get the idea for this?
I had a similar idea, but I had a specific woman in mind and I don't have the hardware yet to make my own lora with her face/body.
Last edited:
Lol, where did you get the idea for this?
I had a similar idea, but I had a specific woman in mind and I don't have the hardware yet to make my own lora with her face/body.
Here's my prompt
(full body view:1.3) Photo of a African American woman who is 45 years old, overweight, with brunette, graying big hair, naked, extremely sweaty, stressed, bruised, dirty
cinematic lighting, vibrant colors, detailed skin texture, beautiful detailed perfect face, intricate sharp details, ultra high res, inside a burning house
I was just wondering what made you come up with the idea of the woman in the burning house?
Oh, sorry.
I sometimes dabble in trauma themes to see how much expression I can get on a woman's face in relation to the surroundings and circumstances.
What upscayler model are you using with Upscayl?
Stable Diffusion 3 — Stability AI
Announcing Stable Diffusion 3 in early preview, our most capable text-to-image model with greatly improved performance in multi-subject prompts, image quality, and spelling abilities.stability.ai
I'm hearing that Stable Diffusion 3 will have a range of different models or configurations, some of which might be too big for home use. Stable Cascade might be better suited for those of us who want to make or use custom models.
Sagging Breast (Breast Realism) Loras
I usually try the sagging at weight of 0.5, but going down as low is 0.1 is viable as well. It's all about making the breasts look more realistic for me.
SDXL:
https://civitai.com/models/164779/legendary-tits-xl
https://civitai.com/models/256298/really-sagging-breasts
SD1.5:
https://civitai.com/models/153636/legendary-tits
https://civitai.com/models/307631/saggy-tits
https://civitai.com/models/310934/saggy-tits-v20
https://civitai.com/models/195920/small-body-and-small-tits
The best loras are still on SD 1.5.
Also, if you are still on SD 1.5 consider installing "Custom-high-res fix" from inside the extensions tab. Combine it with a detailer for better results.
Also I've been getting better results with Upscayl than using automatic1111's internal scaler for some reason. I'm wondering why that is.
I tried RealErsgan and Recrami, this thing is extremely slow. I've yet to try UltraSharp tho.
I've been using Topaz Photo AI, it's pretty damn good and quick.
Upscayl - Nomo
I use Nomo because it hides some of the imperfections of SD1.5, but NMKD Siax might actually be better for SDXL. I would also try the orignal NMKD, superscale.
https://github.com/upscayl/custom-models
They do something to the models to make them work better, that's why you have to get the custom versions.
I've run upscayl on a 1660ti, Rx 6600, and an RTX 2060, each at good speeds. Even the AMD card ran decently. I figure Upscayl is likely using the wrong GPU the reason why it's going so slow for you. Do you have integrated graphics?
If you run it once and then go to settings > logs, you should be able to get the right GPUId.
Last edited:
Yes I'm running the correct GPU, have the setting to 1. It takes like under 2 mins to run one I think it took. I'll check it again.
Just a lot slower than when I ran Topaz Photo. I know they're two completely different platforms. So I guess I should compare the time to A1111, Fooocus or ComfyUI. Running a batch takes like over 30 mins.
But I'll dabble a bit more
@Jake Reed
so I tried a bunch of different models, UltraSharp, Nomo, Remacri (I like the best), tried upscaling a 1280x1280 to 2x and to 1x, all took 2mins, I timed them.
Maybe my input files are too large? Were you running smaller images?
I just did a 1024 and that took 1min and 12 secs
So I suppose it's just the file sizes I was doing that were slower. Will output smaller files and use it.
so I tried a bunch of different models, UltraSharp, Nomo, Remacri (I like the best), tried upscaling a 1280x1280 to 2x and to 1x, all took 2mins, I timed them.
Maybe my input files are too large? Were you running smaller images?
I just did a 1024 and that took 1min and 12 secs
So I suppose it's just the file sizes I was doing that were slower. Will output smaller files and use it.
Similar threads
