App: “Prisma“

A look at how the Prisma app actually works 

Words: Tom Guise
Photo: Getty Images

The Red Bulletin talks to the human mind behind the computer brain replicating creative genius
Alexey Moiseenkov
Alexey Moiseenkov, 26
Alexey Moiseenkov, 26

Best app of 2016? Sorry, Pokémon Go, that accolade goes to Prisma – the photo-manipulating sensation that seemingly transforms your snaps into images in the style of Picasso, Mondrian or Van Gogh.

In truth, it’s the handiwork of this Russian computer graduate or, more accurately, an artificial intelligence he and his team borrowed from someone else – an open-source project named DeepArt

“I thought, ‘How can we connect this idea with people?’ The key was speed.” So Moiseenkov made the rendering process 1,000 times faster and the Instagram generation took notice.

Read more

THE RED BULLETIN: Do you consider what Prisma creates to be art?

ALEXEY MOISEENKOV: No, real art depends on your imagination. Prisma uses deep learning, a neural network, to find all the details on your photo – lines, shapes, circles – then it grades them and completely redraws the image from scratch. It allows you to choose a style, though not create your own. However, we plan to solve this challenge and allow people to create their own styles in the future. Then it will be closer to real art.

Are you surprised at the enormous popularity of the app?

It’s crazy. Only games normally have this speed of growth, and we didn’t spend a penny on marketing – it’s almost completely organic. It’s a unique experience watching how Prisma is growing around the world. In summer, we processed around three billion images in four months. Now it’s around 50 million a day. It all depends on the day, of course. On Saturday, we have more; on a Monday, not so much.

“We want people to explore new styles every day, to keep them interested in creating different images. So we replace some and then see how it goes”
The filter selection is always changing

How do you make money? There are no ads or in-app purchases…

I’m confident we don’t need direct monetisation, so the app is free. We want to monetise from brands, who can create their own style. We already have some [effects] by Gett, the rival of Uber. It’s not a typical ad, which is created by the brand for you; users create the ad for the brand. There’s no logo, no sign, but it’s an ad. People like ads more, and share them, when they’re native to the interface – Instagram, for example. It’s a very powerful way. 

“We improve the neural network based on images people are using, but not in real-time. Once a week, maybe. Make a picture in August and one in September will not be the same”
Prisma’s intelligence evolves

via @sebnero and #prisma #castle #fog #eltzcastle #germany #🏰

A post shared by Prisma (@prisma) on

Read more

What’s the best application of Prisma you’ve seen?

Some people from Bangladesh have created learning books for children in a school. It’s a very interesting case, helping people to learn in different countries, doing some good for the world. I think that’s the way [forward].  

Something really cool is coming... #prisma #prismaart

A post shared by Prisma (@prisma) on

What’s next? 

Creating animation in real-time. Prisma video is already happening on the smartphone, and you don’t need the internet to do it. Our idea is to move all the deep-learning power onto the phone itself, not from a server in the cloud where these supercomputers are calculating the stuff. The user experience will be unified for all people around the world. You’ll be wearing the real AI in your pocket. That’s the challenge. It’s going to be very hard work.

Read more
01 2017 The Red Bulletin

Next story