Fake ASMR Fakes - Fakes of ASMRtists
- Thread starter BlueStripes
- Start date
-
Welcome to the Fakes / AI / Deepfakes category
When posting AI generated content please include prompts used where possible.
Maimy: Includes remakes and alternative versions.
I've been experimenting with a few new things and figured it's worth sharing the knowledge with everyone else 🙂 see my previous guide a few pages back if you're looking for stable diffusion basics.
Expanding images;
As outlined in my previous guide, it's relatively simple for us to nudify someone in stable diffusion by removing their bra or panties, but what if we wanted to take for example a waist-up image and convert it into a legs spread pic or booty shot? Well, it's entirely possible with some extra effort, and gives us a whole new realm of possibilities.
First, take your base image and extend it in the editor of your choice, filling the extension area with white;
Next, throw it into the inpaint tab in SD, and crank up denoising strength to 100%, and CFG to 15-20+. You need to be much more detailed with your prompt, we can't get away with just "naked woman" for this to work; what kind of shot is it (medium close up, full body..), are they facing the camera or facing away, what is their body shape, what pose are they in, etc.
Generating in small batches, you can tell when your settings are "good enough" when it's starting to give you images in the ballpark of what you're looking for. Now you need to generate a big batch (like 10x8), you're going to need to go through a lot of images to find one that matches their body shape. Once you do, you can refine from there as per usual.
Upscaling images;
You can upscale your final results in the 'Extras' tab. Note that you don't want to use an upscaler that changes the image content, and definitely don't want to regenerate faces, as you'll lose the original face. I find that ScuNET and Lanczos are best for maintaining the integrity of the original image. Here are the upscaling settings that I've been using;
Also, the model from my original post has a successor, aEros;
Another one that I've been using is URPM;
Expanding images;
As outlined in my previous guide, it's relatively simple for us to nudify someone in stable diffusion by removing their bra or panties, but what if we wanted to take for example a waist-up image and convert it into a legs spread pic or booty shot? Well, it's entirely possible with some extra effort, and gives us a whole new realm of possibilities.
First, take your base image and extend it in the editor of your choice, filling the extension area with white;
Next, throw it into the inpaint tab in SD, and crank up denoising strength to 100%, and CFG to 15-20+. You need to be much more detailed with your prompt, we can't get away with just "naked woman" for this to work; what kind of shot is it (medium close up, full body..), are they facing the camera or facing away, what is their body shape, what pose are they in, etc.
Generating in small batches, you can tell when your settings are "good enough" when it's starting to give you images in the ballpark of what you're looking for. Now you need to generate a big batch (like 10x8), you're going to need to go through a lot of images to find one that matches their body shape. Once you do, you can refine from there as per usual.
Upscaling images;
You can upscale your final results in the 'Extras' tab. Note that you don't want to use an upscaler that changes the image content, and definitely don't want to regenerate faces, as you'll lose the original face. I find that ScuNET and Lanczos are best for maintaining the integrity of the original image. Here are the upscaling settings that I've been using;
Also, the model from my original post has a successor, aEros;
Another one that I've been using is URPM;
Last edited:
I'm still learning how to do this stuff...But here is some ASMR Glow and one of TipToe
Stable diffusion tips are welcome so I can get better haha
Stable diffusion tips are welcome so I can get better haha
Here you go
And if you want to make some yourself
Last edited: