You have to use your imagination when you look at Matt Richardson’s “photos.” Instead of an image, you’ll see a short description of one: “This is a faded picture of a dilapidated building. It seems to be run-down and in need of repairs.” Richardson’s text-only camera, invented for an NYU class, is an attempt to overcome the limitations of modern cameras — and it’s also art.

The descriptive camera describes what can be seen through its viewfinder. (Matt Richardson/Courtesy of the artist)

Richardson had the idea for the descriptive camera when he was thinking about problems in how cameras capture metadata, the information that is embedded in digital images. While metadata can tell us what kind of camera was used and what time it was taken, it can’t tell us what’s actually in the image, unless someone manually adds keywords and captions. Richardson envisioned a camera that would automatically describe the scene in a camera’s viewfinder, which, when the image was uploaded, would make it easier to find.

Description printout “photos’ from the camera. (Matt Richardson/Courtesy of the artist)

What surprised Richardson about the project was how the descriptions were almost poetic in their banality. His favorite one: “Looks like a cupboard which is ugly and old having name plates on it with a study lamp attached to it,” for an image of, well, precisely that.

“This one caught me by surprise. It was the first time someone made a qualitative judgement about something in the picture,” he said via e-mail. “Not only that, but the person used the word ‘cupboard,’ which you don't typically hear in the U.S. It was the first one that made me think of how much people's backgrounds factor into how they might describe a photo.”

The sparse, stripped-down words can elevate the descriptions to free-form poetry or haiku. Though they are given a few guidelines (use English, don’t be mean when describing people’s faces), the workers aren’t told why they’re describing the photos.

“In a sense, I think of them as creative contributors to the project, whether or not they realize it,” said Richardson. “I think it would be interesting to take photos of people in a heightened emotional state, to see if the workers would try to explain why they're happy, sad, angry, and so on. I like to see what assumptions people make when they see the photo.”

The sometimes-dubious art of Instagram photos has also inspired a similar text photo project: Text-Only Instagram, a parody Twitter account that points out the cliched photography that the photo filtering and sharing service often inspires. Started by Josh Helfferich and Natalie Viscuso, Text-Only Instagram describes the snapshots that Instagram users make when they think they’re being artsy, and it, too, can be poetic in its simplicity.

Portrait of a dandelion with a suspiciously shallow depth of field for a cell phone

— Text-Only Instagram (@textinstagram) April 26, 2012

Latte with foam shaped like a heart.

— Text-Only Instagram (@textinstagram) April 25, 2012

A similar parody account, PicturelessPinterest, makes fun of that site’s own cliches.

The timing of the two descriptive “cameras” is a coincidence, according to a Mashable interview with Helfferich.

“Believe it or not, I actually love Instagram. … I see Instagram as a challenge to create miniature works of art with limited equipment, but there are just some photos that are too easy or too tempting to create with the built-in filters,” he said to Mashable.

So, will we see descriptive features in real cameras in the marketplace soon? Perhaps — but they wouldn’t look much like Richardson’s. For one thing, his isn’t wireless yet, but he says that’s the next step. He’s also considering a Braille printer or text-to-speech addition, to make the camera useful for blind people.

“I don't think it's practical to use a service like Mechanical Turk to append descriptive data to photos, but  I think soon the camera itself will do the work,” he said. “While I'm not well-versed in the algorithmic methods of interpreting pictures, I think it's amazing that computers can already identify faces. Detecting other things in an image can't be too far off.”