An iPhone photo of a bride posted on Instagram at the beginning of November started gaining traction on social media due to its weird nature (via PetaPixel). In the picture, you can see Tessa Coates, co-host of Nobody Panic, trying a wedding dress. The photo is taken from behind, and you can see her reflection in front and at her left, but there’s something wrong: her arms look very different in the mirrors.
Is that a Matrix glitch we thought would appear right before our eyes? Maybe. Coates wrote on her Instagram: “I went wedding dress shopping, and the fabric of reality crumbled. This is a real photo, not photoshopped, not a pano, not a Live Photo. If you can’t see the problem, please keep looking, and then you won’t be able to unsee it.”
The reality is a bit more tricky – but easy to explain. Coates went to an Apple Store to understand why this happened, and it’s all about computation photography.
What’s computational photography, and how does it affect the photo of a bride
“One in a million,” said an Apple Store genius. He explained that the “iPhone is not a camera, it’s a computer,” and an AI decision “stitched those two photos together.”
Basically, when you press the shutter of your iPhone camera, it bursts several images and then combines them in the best photo possible. Since she was moving, the iPhone didn’t consider it was two mirrors, but three different people.
With that, her iPhone captured what it considered the best take for three people – which is bizarre.
Interestingly, the Apple Store employee told her Apple is testing a similar feature to what Google brought out with the Pixel 8 by taking multiple photos and choosing the best ones.
For last year’s iPhone 14 Pro, Apple said the A16 Bionic chip camera hardware performed up to 4 trillion operations per photo. Technologies like Neural Engine, Photonic Engine, and Smart HDR are part of the computational photography umbrella.
That said, you can try mimicking this photo with a few mirrors, as it doesn’t seem Apple will ever fix this “bug” unless the company can improve the algorithm to understand there’s only one person in the picture.
