One of the most talked about developments in the photography and art world is the growing popularity of software-based image generators powered by artificial intelligence (AI). The most popular of these automated AI image generators are Midjourney and DALL-E, both of which are available online with some features offered for free.
To use Midjourney or DALL-E (and the half dozen or so similar products out there), all you need to do is type a short description of what you want the software to create into a prompt box. For instance, write the phrase “snowy winter landscape with deer walking across leaving tracks in the snow,” wait a few seconds, and the software will spit out four different artistic interpretations of what you wrote. Some will likely be more illustrative while others will be photorealistic. Others will be downright surreal.
If you want even more accurate results, you can upload an image, such as a photo or illustration, and the software will “interpret” the shot with its own AI-generated version. While some of the renderings can be pretty awful, others are surprisingly good. In fact, one such AI-produced illustration even recently won an art competition, much to the consternation of other artists.
If you feel a bit creeped out by all of this, you’re not alone. One photographer, however, who is intrigued by the potential of the software is Zach Sutton of Lensrentals, who is also a contributor to Digital Photo Pro. Sutton recently decided to put Midjourney and DALL-E to the test by writing descriptions of some of his photos into their prompt boxes and seeing how close they could render the original images.
Sutton wrote about the experience recently on Lensrentals’ blog in a provocative post titled “Is AI Art Generation Going to Destroy Art as We Know It?”
“With Midjourney and DALL-E being the two vastly more popular AI generation tools, I decided I’d test each of them by using a random selection of images from my portfolio,” Sutton explained.
“Pulling six images from various bodies of my work from over the years, I gave both DALL -E and Midjourney a series of phrases describing the work they would mimic. From there, I selected the best render of the four generated from the AI generation tools. This should give you a basic idea of what their art looks like, and what capabilities they offer. The results of those six images and prompts are below.”
Without giving too much away (you should really read his original post to see all his results), Sutton came away impressed with how well the two AI image generators did despite some caveats.
“While the art speaks for itself, I did find myself really impressed with some of the renders, particularly the work created by DALL-E,” he wrote. “Some of the work, by giving it just a few phrases came out amazingly accurate. Both subjects do seem to struggle with eyes and lips, but to create art as quickly as it does is certainly impressive.”
