One of the latest buzzwords in tech and editing has been AI. While the earliest implementations were a bit of a gimmick, powerful tools and developments from companies like Adobe, NVIDIA, and Luminar have brought AI to the end-user in a meaningful way. Photoshop’s March 2021 update introduces a new AI tool that promises massive resolution improvements for any camera. Does it live up to the hype?
If you’re not caught up on Photoshop’s notes, Adobe Camera Raw has received a new feature called Super Resolution. Currently available in Camera Raw 13.2 and coming soon to Lightroom and Lightroom Classic, Super Resolution uses a machine learning model to “intelligently enlarge photos while maintaining clean edges and preserving important details.” In practice, it’s a one-click way to quadruple your photo’s megapixel count, while retaining much more detail than “dumb” upscaling, like bicubic and nearest neighbor methods.
Photoshop Super Resolution Versus Topaz Gigapixel AI
This isn’t a new idea, however. AI upscaling has been around for a while. It’s even been implemented into existing consumer products, most notably in Topaz’s Gigapixel AI software. Gigapixel is built around a similar principle: train a machine learning model with a set of low- and high-resolution image pairs so that the computer eventually learns what a low-resolution area could look like in high resolution. This model can then up-sample the photos and “create” details to fill in the blanks. Depending on how this is implemented, it can be pretty computationally intense, relying on your computer’s GPU to perform a lot of work.
While there are some differences between how these programs work, with PS’s feature integrated into ACR instead of being a standalone program and Gigapixel offering some more options for customizing the processing, the end results are perfectly comparable.
For these tests, I wanted to take a look at a couple of different types of images that I often shoot and that often benefit from more resolution. To compare, I grabbed some raw files from my Mavic Air 2 (to represent aerial shots) and my Nikon Z 7 (representing architecture and product photography, as well as higher resolution). While these files are unprocessed, each program handled them slightly differently, the most important distinction being ACR automatically applying lens corrections. This resulted in a slight difference between FOV and brightness between the files, but I’m not really considering that relevant in the comparison, as you could pass a processed file through Gigapixel without a meaningful difference. Also, in the following images, the Photoshop Super Resolution version will be on the left, with Gigapixel's on the right.
Increasing Resolution of Drone Photos
In my mind, this is the worst-case scenario for upscaling. While the Mavic Air 2’s files are very impressive for a camera that can fly at 40 mph, they aren’t gorgeous at a pixel level. They can be a bit noisy even at low ISOs, and the Quad Bayer sensor, like Fuji’s X-Trans, has historically had problems with some demosaicing processes.
One thing that stuck out to me when reading about Super Resolution was how it included ACR’s Enhance Details processing step by default. Enhance Details was an earlier foray into ML-powered tools and offered a way to demosaic raw files with fewer resulting artifacts. It’s a very minor improvement in many cases, but I’ve found that it can help in cases of tricky moire, or with atypical sensor setups like X-Trans or Quad Bayer. As a result, I don’t do it by default but appreciate that it’s available.
This combination of improved image quality and increased resolution make Super Resolution seem like a very promising option for use with drones, and I can say it really delivers.
First, let’s talk about processing time and workflow. Loading the raw file into Photoshop, then right-clicking the image and selecting Enhance brings up the relevant menu. From here, a preview is quickly generated, and a new DNG can be created. Via this flow, you still have access to the same features you’d have if you’re processing the raw file normally and can also quickly see what benefit Super Resolution will offer.
With Gigapixel, loading the raw file and setting things up is quite a bit slower. There’s a delay as the preview generates, a significant delay each time you scroll or change an option as it redraws, and finally a very significant difference in actual processing times. Super Resolution produced a finished file in 3 seconds, while Gigapixel AI took 1 minute and 23 seconds.
As for the finished files, Photoshop’s version is significantly better. Two major improvements are visible. The first is an area that has been a problem for many other software tools when dealing with Quad Bayer or X-Trans files: “wormy” looking green areas. In Gigapixel’s version, there’s a very watercolor-y, unnatural look to this area of foliage.
The second major improvement is the relative absence of major artifacts in Photoshop’s version. To personify it, Gigapixel is overly aggressive in “making up” details. It creates faint patterns in areas that should be plain texture and generates noticeable artifacts in areas like text and faces. Photoshop, meanwhile, appears to just deliver a very good upscale. The drone shot, after processing, becomes a 48-megapixel shot. While it isn’t going to match a DSLR for microcontrast and sharpness, it’s surprisingly close and a drastic improvement from the original 12-megapiel shot.
The Best Option for Upscaling Architectural Images
While my Z 7 offers great resolution with its 45-megapixel sensor, more is always better. To that end, I was curious how these two scaling methods would work with a file offering a mix of organic shapes and straight lines, along with some fine details.
From this test file, I observed a similar pattern in usability, but to an even greater degree. Photoshop rendered a finished file in 6 seconds, while Gigapixel took 5 minutes and 1 second to finish its version.
Comparing the two files, Photoshop again delivered a surprisingly neutral file. There are no major problem areas, and the files still have quite a bit of “bite” at the pixel level. As Photoshop auto-applies lens corrections, the FOV is a bit different, but I think these corrections would need to be applied anyway to Gigapixel’s file, as there’s noticeable distortion present in the buildings. At the pixel level, PS’s version only has a slight issue with some fine details, like the stars on the flags. Photoshop renders them as stars, but with a bit of false color creeping in. In Gigapixel’s version, these are unrecognizable smudges as well as artifacts from false color.
Gigapixel also runs into that watercolor problem again along the chain-link fence. Here, Photoshop renders the fence as expected, while Gigapixel’s version is smeary, with individual strands of the fence almost seeming to blur out of focus.
In the architectural details, both are competent. Photoshop seems to err on the side of preserving a bit more noise and texture, while Gigapixel smooths things to a greater extent, but I think you could push either file to the same place with a bit of sharpening and noise reduction.
Conclusion
For about $100, I just can’t see the value in Topaz’s Gigapixel AI product for my workflow now that Adobe’s Super Resolution is available. In my testing across the range of subjects I shoot, Super Resolution delivered equal or better results in every case. Architecture, landscapes, nightscapes, product photos, aerial shots, and more all came out better in Super Resolution. That isn’t even considering the significant workflow benefits: Super Resolution is built-in to Photoshop, respects the existing ACR workflow better, and is anywhere from 20 to 50 times faster to process. If you haven’t tried out Super Resolution yet, definitely give it a try!
May I suggest including a caption below the comparison images or a tag on the image for future articles?. Without mentioning it explicitly it's kind of confusing to tell if we are seeing before/after images, Pp/Topaz, or Topaz/Pp
Thanks, unfortunately with that many images it’s a bit much. Instead, “Also, in the following images, the Photoshop Super Resolution version will be on the left, with Gigapixel's on the right.”
Typing "Photoshop" onto one image, copying, pasting it to five more, then typing "Gigapxel AI" onto another image, copying, and pasting it to five more is "a bit much"?
Considering kindergarteners can remember which is left and right, I was confident you could too.
In the Interests of a professional layout, it would probably be best to caption or label the images. As a student engineer, I am constantly reminded to label everything!
Yeah, it’s always a balance between information density and readability - the inline image sizes are so small that I prioritized getting as much of the screenshot in there as possible. If these were blueprints, it might be a different story :)
I welcome greater competition in the space. Topaz and Adobe can keep trying to outdo each other, and every time they do: photographers win!
Im a bit sour as i just bought gigapixel a few weeks ago...but until this is completely integrated into Lightroom ill keep using Gigapixel for my a7siii Files. It pretty much solves the one weakness of the hybrid camera.
The video-version is mindblowingly good tho. I recently remastered an old documentary dvd with this and it brings the mostly 720p footage with great detail to life. Amazing. Cant wait, until Resolve integrates something similar in the base program...but i guess that will take a few years...
Don't worry, you made a good choice when you purchased Gigapixel AI. In a nutshell, I tried both Adobe and Gigapixel AI on a TIF of a portrait and the Gigapixel AI did a better job. Please see my post below for more details.
I don’t quite think this is a fully fair comparison. ACR is absolutely going to have a better raw processing algorithm. That is a different question then which has a better super resolution algorithm. I would very much like to see a comparison of a TIFF file out of ACR that is then put through gigapixel. I did a quick one-off test myself and was actually happier with gigapixel. Adobe Super resolution had fewer but nonzero artifacts but topaz did a much better job reducing noise and putting in realistic sharp detail. In the end I think I’m going to try both products whenever I’m generating a high-resolution file for output but I suspect I’ll continue to use topaz. In my workflow increasing resolution is the second-to-final step (last being output sharpening with Nik). Creating an eight times larger DNG file and then doing all of my Photoshop work on a file with four times the pixels is significantly slower workflow than finishing the project and kicking off gigapixel while I have a cup of coffee or run a batch before printing. If you think gigapixel is slow wait till you actually wait for the large format printer. Plus I’m always going to have a large camera ready TIFF file; I don’t also want an enormous DNG that needs to be backed up as well.
You are correct, a RAW file should not be processed with Gigapixel AI. I compared both using a TIF and Gigapixel AI did a better job.
I highly recommend trying a TIF vs. TIF using Gigapixel and Adobe.
With enhance details being built in, super resolution really benefits from a raw file. Even giving Gigapixel a finished tiff doesn’t correct its tendency to induce artifacts.
I just took my own advice and tried a TIF of a portrait I processed in Capture One. At first glance they are similar, but when looking closer, the Adobe file has much more artifacting and the Gigapixel AI version was very smooth and natural (while still maintaining detail and sharpness). Plus, Gigapixel AI gives you more options for controlling the processing. I used the Standard mode, but there are also modes for Architecture, Art and Compressed files. On top of that, Gigapixel AI lets you go larger. The Adobe result is good, but the Gigapixel AI output is very good. The trick is to process raw files first with your normal raw processing software, then use the Gigapixel software to enlarge.
GPAI can produce better images - particularly with high-frequency man-made or architectural objects. In fact sometimes it's nothing short of astonishing. You do need to tweak the settings, but of course ACR has no settings at all so you get what you are given. For low-frequency stuff like portraits, GPAI is ever so slightly better, but generally not worth the wait and you have to look really closely.
I made some 30" x 20" prints from a Leica D-Lux (107) 12MP camera and the results were superb, but GPAI has a slight edge.
it seems to work really on low res files too - a quick test on an old Canon D60 file was amazing. New lease of life for old images.
GPAI is worth the fee IMO - I bought a copy after my tests using the demo, but it's hard to fit into an efficient workflow. It's sloooooow, even on my 10 core iMac. But, to my eye at least, it can squeeze out a bit more definition.
Update: I should add that I do not see the change in contrast that the all the sample images above show. Maybe the author is doing something different. I exported my 12MP DNG files to 16 bit TIFF before running them through GPAI, if you run the same DNG file through GPAI, I got muddy results too.
The Topaz apps are GPU driven, so depending on how old your system is, and how much VRAM you have, it may run slower if you don't have a recent GPU and sufficient VRAM.
Using a max-spec current model iMac with 72GB ram and 16GB VRAM, and still GPAI takes about 10-20 times longer to process a file than ACR. It can be worth the wait though!
Here are screenshots of what I got with Adobe and Gigapixel AI. The Adobe image has the artifacting. Be sure to click on an image to really see the differences! These are created from a 16bit TIF file and NOT the RAW file.
Interesting. Also might be influenced by the choice of data used in the models - Giga might skew towards portraits. I’ll run a Tiff through and post a comparison of a different subject.
Why does the GAI image have CA? Also, can you work offline with the latest version? I may be wrong but it appears the app "phones home" to tap into a larger database in order to process, which may partly be the reason for the longer processing times.
With a tiff, Giga is still significantly more artifacted and overly smoothed versus PS.
Something does not seem right. I have used Gigapixel AI previously and it never showed huge artifacting like in your photo above. Did you use the Architectural mode?
Nope, default settings on Giga were used. The same issues are present with either model (standard or architectural), although they may manifest slightly differently.
I believe that the main use of software like this, be it from Topaz Labs or Adobe, is the possibility to enlarge certain segments of an image and still be usable, instead of enlarging the whole image for the sake of printing a XXL sheet. Knowing which part of an image you want to enlarge, first cropping and then running it through the enlargement will also greatly improve the processing speed. If you don't have a somewhat reasonable computer the resulting filesize of uncropped enlarged images can bring photoshop down to it's knees. Since most people who look at a print hanging on a wall rarely use a magnifiying glass to check for artefacts, I think you will find this method yield very pleasing results, especially if you run the cropped image through another piece of sharpening software, like I did on the example posted below. The large image is stitched from two photos (24MP each), I then made a crop of the altar, used the new ACR-enhance feature and afterward ran it through a sharpening process. The result is on the right, with some perspecitve, exposure, and color adjustments.
Which of the two (PSR or GPAI) would work best for automotive light painting images taken with an EM1 mkII for printing at 30x40 inches or larger? Currently my first step is run RAW through DXO 4 for Deep Prime noise reduction and optical corrections, output as DNG. I then load into PS as layers and mask multiple images to create the final result, then flatten image. After flatten I go to LR for final global edits. Next step would be PSR or GPAI for upsizing, then output sharpen. Thoughts?
My thoughts -- if your images show a lot of noise - reduce noise before resizing. But generally I allways apply any resizing process before doing any other post-processing. So in your case, depending on the amount of noise, either at a first step or as a second step after your deep prime noise reduction. But just do two test runs: one using either software at the beginning or the end -- I kind of think you won´t really be able to tell a difference, at least not once the image is printed out. Since a 40x30 inch image will be viewed at some distance (propably about at least 40 inches away), the subtle differences that might show from "pixel-peeping" won´t be perceived. Since you actually asked for which software is better suited - I use PSR since its faster. And even though GPAI -- like all other Topa AI software,-- has more controls, most people still just use the "auto" function" and leave it at that. Good luck!
In addition to my TIF portrait samples posted in this comment section, I tried an architectural JPG shot. My results were the same: Gigapixel AI does a better job compared to Adobe. The photos speak for themselves. Be sure to click to see the photos large! On a technical note, I used the Architecture mode when using Gigapixel AI and did not make any other adjustments.
In the end I think people are just going to reach for the immediate solution in their RAW processor. The good thing about Adobe adding this feature is that Topaz will be force to drop the price of Gigapixel, which frankly is overpriced, so it'll be win-win all around.
I love AI Gigapixel but it's so slow that by the time the uprezzing is done I've missed my client's deadline. Further, if I am correct Topaz has a subscription model for updates which for some products is as much as the entire Adobe CC price. So for me this new Adobe feature is a real plus. Hopefully to compete Topaz will work to speed up AI Gigapixel and offer the updates free as they used to or at a significantly more modest price than they charge now, as I would like to have both options at hand for critical work.
This comparison is a little off. The automatically applying lens corrections and camera raw processing is making the Photoshop images look better in each comparison. I believe Dan Donovan is correct in having the challenge run on a 16-bit TIFF would be a more fair comparison.
If Giga is working with a tiff, the same pixel level problems and artifacts that I discuss are still present.
I considered purchasing Topaz Gigapixel AI about two months ago. I wanted to enlarge a few bird images by 2x I realize this does not add uncaptured detail. I downloaded the trial version. I was literally floored at how many artifacts I saw when comparing 1-to-1 with the original. Gigapixel just didn't multiply the number of pixels, it literally manufactured all kinds of little patterns not present in the original, and they were detectable even when the resulting image was viewed at less than 100%. I tried the regular PS bi-cubic sharpener as a comparison and it did a much better job of doubling the image while keeping it true to the original.
I like some Topaz products like Denoise AI and Mask AI (though I'm not sure how AI is really occurring).
It sounds like we may have great upgrade with Super Resolution.
I tried it too and not only is it slow as a damn turtle, I did not see any additional detail that it supposedly adds. It honestly looked like garbage in every single case. It wasn't even "meh this is good but not good enough to buy" it was "this is destructive to the image" - I always ended up just upsizing the photo in PS instead.
The ACR version works fantastically - not only very fast, but actually increases the image quality. And it's free, since I have Adobe.
The reality, though, and I say this as someone who owns a large format printer and regularly uses it, extremely high resolution is overvalued for printing unless you're 1) printing very large from, say, a 12MP image - though even then I've printed many of my D700 images that look beautiful without any upsizing, or 2) you crop heavily, which is probably the most valuable use of it. I've absolutely never needed to upsize a file from my Z7, unless it was heavily cropped.
Of course, a small percent of people may be doing special use kind of printing, but most are not.
There is, however, value if the process is actually able to enhance detail. Which Gigapixel absolutely did not for a single file I tried. And before someone comments about it, yes, I was using TIFF files, not RAW.
Thanks for taking the time to write this up, as I didn't even catch the implementation of this myself. I do look forward to more nuanced comparisons, as many people have rightfully pointed out, there is more to this than simply processing for raw files.
Use cases vary widely, and its great that there are tools out there to handle different things, some will naturally be better than others, but not all the time necessarily. I sometimes enlarge old scanned photos that are in jpg or tif. In these scenarios that i tested, it took little more than a few minutes to see Gigapixel results flexed easily over ACR, in terms of sharpness, detail and noise suppression. Not to mention, Super res feature is strictly set on x2 scale (for now?), Topaz offers greater flexibility, of course like anything it has its limits.
Compare this with raw camera data for input as you have, then yeah that may be a different story - in that regard super resolution looks impressive and am excited to see how it continues to compete.
I've been using gigapixel for over a year. I do large format printing. This software is a game changer
when it comes to increasing file size. As with any upsampling software, the larger the original file, the better
the results. The new ACR is a good start for Adobe, but is not in the gigapixel class. Not even close.
Gigapixel does 2 important things that ACR doesn't do. It reduces noise to very acceptable levels. It also
sharpens the image as it upsamples. I use many image editing softwares. Gigapixel is is great.
Haven't tried it yet but definitely will. AI functions are everywhere nowadays and I'm really curious to see how are these features different from the ones in Photoworks or other editors. I loved the way one of the previous posters had put it: photo editors try to outshine each other and photographers are the ones who definitely win.
I find that all the ‘before’ images are very low contrast. Muddy.