[Tech]How AI Works in Prisma App

Millions of people are sharing their Prisma creations on social networks like Facebook, Instagram and Twitter. Why is it so popular? One of the reasons can be its attractive filter style. Prisma is by no means the first app to create wonderful, fantastical, and artistic photos using filters, but most of the users agree that it does it the best.



How come? AI (artificial intelligence) helps a lot. Differed from the other apps that merely apply light and color filters to images, Prisma app uses a combination of neural networks and artificial intelligence to take the art styles of famous artistic effect like Van Gogh, Pablo Picasso and Edvard Munch as well as other art styles, and then combines them with your photos to create images that only referring to the original picture for guidance in the creation of the completely new image instead of making changes on the original photo or adding a layer above it to enhances the visual appearance of the image.

The Prisma app, produced by a four-member Moscow team, uses a deep learning method named convolutional neural networks. It was inspired by an online algorithm for converting photos into paintings. The similar apps including DeepArt.io from Germany, Ostagram from Russia and Masquerade, a Belarussian startup which was acquired by Facebook, utilizes a self-learning algorithm for altering the faces of users in images or videos. The app, however, is limited to fitting effects to various facial shapes.



"We're not just overlaying like an Instagram filter. We create the photo from scratch. So there is no photo, we took your photo, then perform some operations and give a new photo to you. So deep learning is like an artist, something like that."

– Alexey Moiseenkov, CEO, Prisma labs inc.

Prisma, unlike these apps, uses Texture networks. Prisma doesn't just apply the chosen filter on the image like in the case of other famous apps, rather it takes the image and analyses it, use the texture and style from a famous artwork. Actually, the artificial intelligence program behind Prisma uses two images to create a third image. The first image is obviously the photo from the user’s device and the second one would be the artistic style. The AI algorithm analyses the input image and intelligently applies textures and effects in this image to create a great result in the exported one. And in this way, the app generates the imagery from scratch like the blank canvas that's why these images look like real-life paintings.