Even with all our emojis and emoticons, a typed message will never carry the same sort of emotion as a face-to-face conversation. Sure video calls have gotten a lot easier these days between FaceTime and Skype calls, but who makes calls anymore?
To help combat the problem, researchers from the University of Cambridge are saying that they have developed the most expressive, human-like avatar ever created. Dubbed, "Zoe", the avatar can read text out-loud with all the emotion a disembodied face would have.
To create this human-faced avatar, the scientists had an actress named Zoë Lister read over 7,000 lines of dialogue while tracking her facial expressions. The data from these exercises were used to train a virtual model. In case you think this is all just a recording, the underlying mesh structure of the virtual face can be manually manipulated for some real freaky expressions.
Zoe basically reads out a message you type in with the emotion you select manually on a bunch of sliders that includes: anger, sadness, happiness, tenderness, fearfulness. In a case study, a group of volunteers correctly recognized the emotion of the virtual avatar (77 percent) at a higher rate than the actual human actress (73 percent). So once again robots have beaten humans.
The team behind Zoe is currently looking into possible applications for the avatar. One of the most prominent uses for this technology could be to help autistic children read emotions and the deaf to read lips. The scientists also believe the technology could have a future in gaming, audio-visual books, and online lectures.
Since the system is practically a template, anyone could record themselves, or someone else, to create their own personalized assistant. The system is also extremely lightweightjust 10-megabytesso it might end up being implemented into our phones and tablets someday.
Can you recognize its emotions? Leave a comment.