our upscaled and enhanced photo of Father Francis Mulcahy (William Christopher) on M*A*S*H
seen from United States

seen from Italy
seen from United States

seen from Malaysia
seen from United States

seen from France
seen from United States

seen from Maldives
seen from United States
seen from Yemen
seen from United States

seen from Malaysia

seen from Malaysia
seen from Yemen

seen from United States
seen from Malaysia

seen from Malaysia

seen from France
seen from United States

seen from United States
our upscaled and enhanced photo of Father Francis Mulcahy (William Christopher) on M*A*S*H
Hi, I saw your post where you talked about the new upscaling program you found - you mentioned spotting artifacts from upscaling and was wondering whether you could elaborate on that? Just out of pure interest, I looked into your Behind the Scenes posts and couldn't find anything related to it, so I thought I'd just ask
With topaz gigapixel AI, the upscaling software I was using before, you would get really obvious artifacts, especially with screen tone, which is this btw:
It's how manga achieve gray tones, even though they only used to get printed in monochromatic black and white.
Topaz often turned screen tone into a gray, blurry mess. It also really struggles with parallel lines and repeating patterns, such as on Laios' armor or the shadows in Yotsuba.
This isn't really Topaz's fault because this simply isn't what the software was made for. Here's an example of how it fares with the kind of picture you're supposed to use it with:
Art by loish
Anyway, MangaJaNai, the new software, still struggles with things like parallel lines, but the artifacts are a lot less obvious.
I really only noticed because I was looking for them. And other than that, the software is doing an amazing Job, especially with screen tone.
Lastly, here are a few examples of more obvious upscaling artifacts with MangaJaNai.
You can see the screen tone pattern warping around drawn lines, and the parallel lines also look strange again.
This is a rather extreme example and the artifacts are only this obvious because, as mentioned in my other post, the original image used solid grays instead of screen tones. So the software basically had to just kind of guess how the screen tone should look like.
A plaza in downtown Denver, Colorado was a “harsh place,” a “terrible concrete plaza,” explained Kasey Toomey, PLA, landscape architect, art
What a change from a concrete plaza to a lush landscaped area.
Kudos to All involved for the transformation of Denver 17th Street Corner Plaza.
Here's a very special one. I've been playing around with the tech and this is one of the most difficult things to do so far. The first image is the original 1024x1024p Bing AI image, raw and unedited (file size 192kb).
The second is using the tech. It's a huge 4096x4096p image, weighing in at a whopping 11mb!
Because of the difficulty in generating this detail, this is the only one of these I'll do for now. But as the tech improves and it becomes more accessible, I'll be able to make some of the best AI-Satin-Chic art ever!
Just LOOK at the difference it makes - here's a comparison of the detail in the head. It's a shame that Tumblr doesn't let you fully appreciate just how much detail is in the new images.
For now, if you really want to see the image in all it's detail, check it out on my DeviantArt.
I am melting
minecraft build idea for an smp im joining soon
Playing around with Gigapixel AI, and of course I’m testing it out with my massive collection of Trent pics.
(Click for full quality...I think)
Photographer: [TBA]