Photorealistic AI Generated Images
- Thread starter SUNBO
- Start date
-
Throughout the month of April 2024, participate in the FileJoker Thread Contest OPEN TO EVERYONE!
From 1st to 30th of April 2024, members can earn cash rewards by posting Filejoker-Exclusive threads in the Direct-Downloads subforums.
There are $1000 in prizes, and the top prize is $450!
For the full rules and how to enter, check out the thread -
Akiba-Online is sponsored by FileJoker.
FileJoker is a required filehost for all new posts and content replies in the Direct Downloads subforums.
Failure to include FileJoker links for Direct Download posts will result in deletion of your posts or worse.
For more information see this thread.
Indeed, and as an early Christmas present, here is a quick remake of your image!
Remake:
Upscaled remake:
That's a really pretty Japanese gnd from AI, would def follow her if she's a real jav actress. Is it possible now to make variations with the same face and body? She could be the first jav star to emerge from here lol.
It might be possible, but would be tricky as there is only one image from the front of her right now.
The easiest for now is to simple just reuse the generated image and inpaint various aspects like this:
I am using my own rig running stable diffusion web ui by AUTOMATIC1111 locally, like I explained here using a stable diffusion 1.5 model. The fastest results is to have a graphics card with lots of ram, at least if you want to generate images of a bit higher resolution, but you can also use various methods of upscaling. The goal is to generate as high resolution as possible, and then possibly use upscale as it usually flattens the image in one way or another and you loose detail.
My graphics card has 12GB RAM, but with the recent nvidia drivers you can render high resolution by offloading CUDA work to system ram, at the cost of slower image generation. You can probably use any supported nvidia graphics card with CUDA cores, even 3GB cards like 1060 I guess? Not sure of the requirements when using other platforms like MacOS or AMD.
Sometimes I also do some minor adjustments and cleanups, mainly using a clone-brush tool or soften/unsharpen tool on certain parts of the image that stands out after inpainting, you can use a free paint software like Krita.
Started this image a while back, decided to finish and make some final touches today!
Would you enjoy a drink at this bar?
Would you enjoy a drink at this bar?
Came out nice! Thanks for sharing!Started this image a while back, decided to finish and make some final touches today!
View attachment 3384193
Would you enjoy a drink at this bar?
Some quick tests with the realistic vision 6.0 beta1 model
Non-upscaled:
Upscaled versions:
Left is using 4x-UltraSharp, Right is using 4x-Foolhardy-Remacri
Remacri gives more textures in general when upscaling instead of a softer "watercolor" and less detailed image,
but it can also mess things up a bit, you can see it in the right eye for example.
There are lots of other upscalers of course, maybe need to investigate a bit more.
I'm also thinking it might be better to do another inpainting pass on eyes and lips after scaling to restore some sharpness and detail.
Upscaling is performed using the A1111/stable diffusion web ui as well if anyone is wondering.
Non-upscaled:
Upscaled versions:
Left is using 4x-UltraSharp, Right is using 4x-Foolhardy-Remacri
Remacri gives more textures in general when upscaling instead of a softer "watercolor" and less detailed image,
but it can also mess things up a bit, you can see it in the right eye for example.
There are lots of other upscalers of course, maybe need to investigate a bit more.
I'm also thinking it might be better to do another inpainting pass on eyes and lips after scaling to restore some sharpness and detail.
Upscaling is performed using the A1111/stable diffusion web ui as well if anyone is wondering.
Last edited:
Yes, it all depends on how much time you are willing to spend.
Here is an example I just did, forgot to specify someone was holding the thing in the prompt though...
![[REAL-564].jpg](https://www.akiba-online.com/data/attachments/3386/3386899-1cd8a9490a9271374288b7b995853153.jpg "[REAL-564].jpg")
You can change or add whatever details you want with inpainting, bonus remix and upscale:
A YouTube channel I highly recommend for AI tools, setup and usage is Nerdy Rodent.
This video (and a follow up) covers Reposer - maintain same character (and clothing in the Plus video) while changing poses:
Segment Anything simplifies masking for inpainting:
Seriously considering playing around with ComfyUI instead of Automatic1111. Looks amazing:
Also has videos covering AI audio, video and other miscellaneous tools.
This video (and a follow up) covers Reposer - maintain same character (and clothing in the Plus video) while changing poses:
Segment Anything simplifies masking for inpainting:
Seriously considering playing around with ComfyUI instead of Automatic1111. Looks amazing:
Also has videos covering AI audio, video and other miscellaneous tools.
Had the hardest time convincing myself she isn't real...
