The Future of Images Is ‘Trust Nothing’
Adobe’s lab is cooking up some wild, AI-powered image-manipulation technology called Project Strike a Pose
Whenever someone snaps my picture, I automatically strike a pose, doing my best to compose my features and limbs into something that doesn’t look awkward or silly. I guess I could just stand there, arms at my side, staring expressionlessly into the lens.
Then an AI engine could take my face, mold it into a rictus grin and reassemble my arms and legs into John Travolta’s Saturday Night Fever dance floor pose.
That’s the world Adobe envisions as engineers in its skunkworks labs take what is apparently the almost limitless power of its AI- and Machine Learning-powered Adobe Sensei to figure out new ways of altering still and moving images long after they’ve been shot.
The news that Adobe is working on next-gen image manipulation came during this week’s annual Adobe MAX conference. It’s where we learned about Photoshop and Illustrator coming to the Web. I thought that was a big deal, but it’s unlikely to have the same kind of societal impact as Project Strike a Pose and Morpheus.
Both still-in-the-labs projects manipulate images in wild and arresting ways. Morpheus extends to video Photoshop’s current Neural filters that let you turn a frown in a photo into a smile. They can even add facial elements like the pornstache you never had.
Project Strike a Pose, however, might be more concerning. It takes full-body photo subjects, uses images of people posed in completely different ways, and applies that pose to the original subject. In the demo, they did this with a woman standing in a park who looks like she’d rather be anywhere else. The developer used a different photo of someone else smiling as she walked through the frame, applied that to the original photo of the woman, and, voilà! the woman’s body was contorted into the same pose, and she now had a smile on her face.