DALL-E 2, an artificial intelligence system that can create photo-realistic images from just a brief descriptionhas been used by a photographer to edit their photos and can sharpen a blurry image.
Nicholas Sherlocksent the DALL-E 2 edited images to YouTuber Michael Widellwho was rightly blown away by the possibilities of the technology and went so far as to question whether the possibility would be the “death of photography”.
The images submitted by Sherlock show an out-of-focus ladybug miraculously sharpened by OpenAI’s software. To restore the image, he erased the blurry part of the ladybug’s body and issued a text prompt that read “Ladybug on a leaf, focus high-resolution stacked macro photo.”
This ingenious method uses DALL-E 2 for something it was not designed for, but it can become a powerful and important tool for photographers.
Speak with PetaPixelSherlock gave another example of a photo he mounted on DALL-E 2, this time of an egret in a ditch.
“Thanks to DALL-E’s inpainting, you can upload an image, erase part of it with a brush, tell DALL-E what to put in that space, and he’ll paint it for you,” he says.
“I erased the egret and erased a space to the right of the image, and asked DALL-E to generate ‘baby elephant baths, wildlife photography’.
“In this case the results cannot be looked at closely, the elephant is a bit too sketchy, but this is not an inherent limitation of the technology and this will improve over time. They look great in miniature format.”
Sherlock points to the baby elephant water reflections generated by DALL-E as a remarkable feat that was “done much better than I could have done”.
Inpainting and Outpainting
DALL-E 2 offers the ability to “inpaint”, where editors can create subjects in an image using only a text command, as demonstrated above.
However, it can also “paint over”. Sherlock gives this example of a tilt-shit photo where he wanted the crop to be “a little looser.”
“I increased the size of the canvas in Photoshop to give it transparent borders and uploaded that image to DALL-E. I say it to fill it in with the prompt ‘A city in the fall, 35mm tilt-shift photography “Velvia.” This corresponds to the original image.”
Sherlock even sent a comparison photo using Photoshop’s Content-Aware-Fill tool, which unsurprisingly failed to create a much wider scene like DALL-E’s.
Whether Adobe can create something like what is possible with DALL-E 2 for the wider photography community remains to be seen.