|Google Offers AI Help To Create Fantasy Creatures|
|Written by Kay Ewbank|
|Friday, 27 November 2020|
Google has released a prototype AI-based tool called Chimera Painter that takes a basic sketch and turns it into a fantasy beast using machine learning.
The tool is aimed at developers who want to create art for video games, according to Andeep Singh Toor, Stadia Software Engineer and Fred Bertsch, Software Engineer, Google Research, Brain Team.
They say that game artistis need not only artistic creativity and technical knowledge, but also have to
"quickly iterate on ideas and produce a high volume of assets, often in the face of tight deadlines. What if artists had a paintbrush that acted less like a tool and more like an assistant? A machine learning model acting as such a paintbrush could reduce the amount of time necessary to create high-quality art without sacrificing artistic choices."
Google's possible solution is Chimera Painter, a trained machine learning (ML) model that automatically creates a fully fleshed out rendering from a user-supplied creature outline. Chimera Painter adds features and textures to a creature outline segmented with body part labels, such as “wings” or “claws”, when the user clicks the “transform” button.
The results look impressive, as can be seen below:
However, our experiments suggest you'll still need quite a lot of artistic skill and to spend more time than we were willing to dedicate in order to get something really good rather than a worrying pantomime horse. It's also quite tricky to think of what wings or horns ought to look like that doesn't end up looking like the samples:
Under the covers, the system was trained using generative adversarial networks (GANs), with feedback supplied by real artists with the aim of creating images of creatures that could be used in a fantasy card game. GANs pair two convolutional neural networks against each other: a generator network to create new images and a discriminator network to determine if these images are samples from the training dataset (in this case, artist-created images) or not.
Google researchers used a conditional GAN, where the generator takes a separate input to guide the image generation process.
Once trained, the models were asked to generate "multi-species chimeras", based on outlines provided by artists. The best performing model was then incorporated into Chimera Painter and outputs beasts like this:
or email your comment to: firstname.lastname@example.org
|Last Updated ( Friday, 27 November 2020 )|