ON THE HOT TOPIC OF AI: Is Adobe’s Photoshop (Beta) Generative AI serious or fun photoart?
Using a single photograph of F1 cars taken during the 2023 Canadian Grand Prix to create AI image examples, this commentary puts Adobe’s new Photoshop (Beta) Generative AI under the microscope for whether it is serious or fun photoart.
Gen AI is becoming a powerful, amazing tool in the photography world as well as elsewhere in society -and that, I point out, conveys huge responsibilities for transparency, copyright and veracity.
As a fine art photographer who has a serious interest in technology, both generative AI and F1 are exciting and fascinating to me and combining both of them as a thought + visual project - whether Adobe’s Photoshop (Beta) Generative AI is serious or fun photoart - just seemed natural.
This is the original image I shot of the F1 cars coming out of the short straight just before the S curve where we were seated at the Canadian Formula 1 Grand Prix in Montréal in June. All AI Generative Fill images below are derived from this one.
Here’s the prompt I used: “make cars appear to drive on water going through tropical forest”. I’m impressed! Gen AI did a great job suggesting reality in an unreal situation, the displacement and splashing of the water create a convincing sense of movement and speed in an entirely different medium of water. The color, leaves and contour on both sides of the forest look realistic. In addition, the image shows one of Gen AI’s strengths - its ability to render nature well. Note the perspective, lighting and scale - excellent. A+
Here’s the prompt I used: “make cars appear to drive on water in a rain storm while going through tropical forest“. I really like the waves, the turbulence and color of the water, the whipping wind - it looks pretty realistic (suspending reality here) assuming the F1 cars could race with full wet tires on water. The Fill changed the forest’s configuration and composition. It doesn’t look as realistic to me as the first image. B+
Then, the prompt shortened: “make cars appear to drive on water.” Pretty cool image, extraneous background removed, water looks calm, you can even see a reflection on the first F1 car and the waves behind cars indicating movement. Even the color and “texture” of the water appear natural. A
The prompt was to make it appear as if the F1 cars were out in space driving through the solar system and the Starship Enterprise was in the background - but I don’t have it verbatim. I think I see the Starship Enterprise in the mid-right border but overall it’s a fun, fantasy of space, although the cars look as if they are floating rather than racing.. I suspect others have discovered a Must Fix in Adobe’s next version that doesn’t exist in Beta - find a way to save the prompt with the image. B+
OOOPS!
On the previous image of F1 cars driving on calm water, I specified this prompt: “add a small alligator's head in front of first car” - and you can see what happened. The poor F1 cars now have to drive through mud (where did that come from?) in the domain of one or two (your choice) giant alligators. No, Adobe, Generative Fill does not “automatically match perspective, lighting and style of images” yet, but it certainly does “enable users to achieve astounding results!” D-
The text prompt to first image with cars driving on water was: “add dinosaur head in tree tops on right.” Whoa! I fear I fried Gen AI’s brain with that request. The water disappeared, replaced by dirt track, my prompt to add a dinosaur was refused, and the F1 cars, especially the lead, look bizarre. D
WHAT’S NOT TO LIKE!
To be very clear, Gen AI is amazing at replacing unwanted elements in images with appropriate content aware imagery just by using the lasso tool to select the target content and leaving the generative fill prompt blank. In short, Adobe’s AI Fill can be remarkably savvy. Once you introduce it to your Photoshop editing repertoire, you will likely use it from then on. You won’t be disappointed.
Further, kudos to Adobe for providing three to six image interpretations with each generative fill. It’s useful to have the optional images which can be similar (rarely) or wildly different.
The text prompt to image on left was: “make each car appear to become a cartoon character in the anime style,” which it did. A couple cars are pretty cute and have some character, the rest are illegible. Gen AI did add a number of extra cars, but, hey, who’s counting! C
THE FUTURE OF ADOBE’S PHOTOSHOP GENERATIVE AI
This is the Beta version after all, and no doubt Adobe has been intensely training its algorithm to refine its content focus and transparency for the official first version.
The merits of basing its Generative Fill algorithm on its Creative Cloud app free stock images “openly licensed content and other public domain content without copyright restrictions”, no doubt, present many more questions than answers and will need to be not translucent but totally transparent. Many photographers’ livelihoods - and trust - depend upon this.