Startup allows users to create realistic animated facial avatar from a single photograph
December 14, 2016 by Bob Yirka
(Tech Xplore)—A team of researchers working at startup Loom.ai, has introduced new technology that allows a user to submit a single image of their face and have it translated into an animated version that can be used as an avatar in virtual reality systems. Company reps recently spoke with the press regarding the new technology, the 1.35 million it has received in seed money and where they believe VR is heading.
Virtual reality is becoming big business, whether it is part of video action games, fantasy adventures, Facebook conversations or even text messages. People want to be able to interact in lifelike ways, Loom.ai CTO Kiran Bhat told onlookers, and that involves injecting "salient features" into virtually recreated faces. The company appears to have achieved that to some degree using machine learning algorithms as evidenced in YouTube videos the company has released that depict avatar versions of several celebrities. Bhat, who previously worked at Lucasfilm, creating facial performances for characters in Pirates of the Caribbean and The Avengers (such as the Hulk), also suggested that the technology for building avatar faces for everyone now exists in products his company is making. Mahesh Ramasubramanian, who previously worked at DreamWorks, has also joined the company as a visual effects artist.
The animated avatar faces the company creates convey emotions that the person behind them is feeling, such as concern, interest or even anger. The set of emotions at this point appears to be constrained to just a small set, however, as they are repeated for all the examples. Still, there is no denying the company has taken a step forward in avatar creation—most current systems allow users to start with a base face and add features from a menu. With the new software from Loom.ai, users need only supply a single front facing photograph—the software does the rest. The result is a 3-D representation that can be viewed from user-chosen angles and, in the future, inserted into a virtual world.
The company will be licensing the technology to other (still unknown) third-party platforms, which will then use what has been created to improve the avatar capabilities of their own software.