Deep Forger is a Twitter account which will generate a version of any picture fed to it in the style of any other picture.
The results can be entrancing. Here’s me at a party in the style of Rembrandt:
And here’s Taylor Swift in the style of Pablo Picasso, which was less successful:
If you feed the bot just one input, it automatically merges it with a great artist. But you can also submit two pictures, and it will generate a version of the first in the style of the second. So you can get your headshot in the style of Japanese iPhone game Neko Atsume:
Or the Firefox logo in the style of the Mozilla logo (bit niche, this one):
Others are … weirder.
And some are just perfect:
Deep Forger is only the latest in a long line of robots doing trippy things with Neural Networks.
First, there was Deep Dream, the Google-built algorithm which purports to let us see what a computer could dream about. The robot used a neural network to identify images by their key features, but it turns out that if you create a feedback loop and feed the same network white noise, the results are pretty trippy:
Then, Google freed its neural networks to run across the internet. The company made the software behind the images open-source, allowing other users to make their own horrifying images:
In September, a team of German researchers managed to build a similar neural network, but with a much more focused goal: mimic great artists.
At the time, Hannah Jane Parkinson wrote that the algorithm could “accurately copy the painting style of artists as disparate as Pablo Picasso and Vincent Van Gogh – and in just 60 minutes.” Deep Forger beats that by fair margin.
The newest bot is created by Vienna-based game developer Alex Champandard, and is open for submissions. Why not try it yourself, and highlight the best in the comments below?
This article was written by Alex Hern, for theguardian.com on Thursday 26th November 2015 12.25 Europe/Londonguardian.co.uk © Guardian News and Media Limited 2010