There’s no question that machine learning and artificial intelligence have radically changed the world over the past decade. From the early days in 2012 when an elite team at Google Brain taught a state-of-the-art computer to recognize a cat to today when anyone with a Chromebook can use Google’s open-source TensorFlow for their projects, machine learning is ubiquitous popping up in all kinds of places. However, one place I didn’t expect to see it was in a web app for making manga portraits.
Giga Manga is an online experimental platform that uses machine learning to turn your scribbles into manga illustrations. It uses a suite of AI and machine learning tools — trained by over 140,000 high-res images — to transform your doodle into the protagonist of a doujinshi.
The process is pretty simple. To start, draw a doodle: a circle, a squiggle, an outline, anything. Google will give you a rough outline of a portrait based on the 140,000 training images. Next, it’s up to us to use our creativity to fill in a few details like lines and colors that Giga Manga can use to complete the rest of the image for us.
The project is part of a partnership between Google, the Japanese Ministry of Economy, Trade, and Industry, and 12 Japanese cultural institutions called Manga: Out of the Box. On its Arts & Culture page, Google showcases the rich lineage of manga from its ancient roots in Chōjū Giga scrolls, to the 19th century when western art discovered Japanese prints, to the post-war birth of Astro Boy, to the thousands of titles available online and on shelves around the world.
Giga Manga is available to play with right now, but there’s no telling how long it will be around, so don’t waste any time if you want to see one of the directions machine learning may be taking us in the future. There is one caveat; you’ll need to have a modicum of art talent, or you’ll end up like me with an AI-generated pile of hot garbage.