Adobe's newest version of Photoshop comes with an intriguing feature in the Camera Raw dialog, Super Resolution, which promises to leverage machine learning trained on millions of images to upsize images to quadruple their original resolution. Does it hold up to its promise? This excellent video takes a look at the new tool and how it performs in the real world.
Coming to you from Brian Matiash, this great video takes a look at Adobe's new Super Resolution feature, which promises to double the number of pixels in both the horizontal and vertical dimensions, thereby quadrupling the overall resolution — a literal "enhance" button. Such a feature has a huge range of potential uses. Wildlife or sports photographers, who often have to significantly crop photos, could use it to recover detail in images. People shooting with low-resolution sensors could use it for producing large prints. Landscape photographers could use it simply to add more detail to a shot. After being trained on millions of images, it seems the tool has some serious potential and could end up being a popular feature for a wide range of photographers. Check out the video above for Matiash's full thoughts on the new feature.
Nice. I can't wait to try this and compare with my copy of Topaz's Gigapixel AI.
Definitely would be a good comparison to see!
Oh it beats Gigapixel AI for sure! I ran some comparison tests upscaling a 36mp raw file. Photoshops "Super Resolution" took about 30 seconds, and provided a lovely clean image with subtle enhancements to minute levels of detail.
Gigapixel on the other hand took 25 freaking minutes to do the same thing! In my case it was 1h15mins, as the software outright crashed on the first two attempts. The result was similar to Photoshop's, but a little more aggressive and produced kinda waxy looking details.
Rock on, Alex! Thanks so much for sharing my vid on Super Resolution! I hope everyone enjoys it!
Oh great, now those that steal lower rez images can bring back the quality... Sheesh
My tests show Gigapixel AI has a tiny edge, but it depends somewhat on what you are ressing up. And GPAI is seriously slow even on my beast-Mac. Super Res works really well on older images - I tried ACR Super Res on an old Canon D60 file and it was amazing!
GPAI just might have the edge but it's USD99 and it's not good in a workflow - having it in ACR, and soon Lightroom, will be a boon.
I shoot with an original 12 megapixel Canon 5D, so this will definitely come in handy when I need to deliver photos bigger than what my camera can make.
D3s user here and I was thinking the same thing..lol
i tried this on a raw and it blew my mind! cant get this to work with a JPEG though? can anyone else help me get this to work on a jpeg ?
You have to export the JPEG file to DNG first. One suggestion is to use the Neural filter to remove the JPEG artifacts before you export.
I was pretty impressed. I like Topax GPAI but by the time its processing is finished my client deadline has passed, so this is a terrific alternative.
Adobe’s game changing AI resolution enhancing technology gives my favorite Lumix S1 and the S5 a real advantage over Nikon and Canon (and some Sony) cameras. Resolution boosting works significantly better with cameras with no anti-aliasing filters. Early samples show that results can be stunning when applied to raw files with no in camera filter. With AI technology improving rapidly, unlocking the potential of out of camera creativity, it is time to give Panasonic cameras serious consideration.
Just upgraded PS. This tool beats Gigapixel AI hands down. Doubling my A7RIII arw raw, took maybe 2 minutes, and I had other apps open. using GAI, EVERYTHING else needed to be shut down and it still took a good 20 minutes for a res-up of no more than 20%.
I’m a C1Pro user and will test if I should res-up using ACR and then import the res’d DNG into C1Pro, or should I first optimize the native arw file in C1Pro, process to a tiff, and then rez the tiff in ACR?
If you bring the Enhanced-DNG file into C1P you might find that it's still the original file size. I suggest working in C1P first, then uprezzing a 16-bit tiff in ACR second. This may not be optimal however, the ACR tech seems to work best on original raw files whilst Gigapixel seems to like 16-bit tiffs best.
Much appreciated Nick! Either way, sure beats the “Genuine Fractals” days!
I’m so excited! I’ll be up-res-ing my pictures from Brazil taken on the 8mp EOS 30D. Yay, big print time!