It is a well-known – if entirely unscientific – fact that the two most powerful forces in the universe are celebrity culture and nostalgia.
We here at Techly haven’t yet ruled out that Ryan Seacrest isn’t an AI composite sent back in time to make the creepy kid from American Beauty look like Orson Welles.
As for nostalgia, there is no greater example than the fact that Vin Diesel’s woefully derivative and cringingly awful Triple X movie franchise is being resuscitated.
With that in mind, it was with no particular surprise that we learned that researchers from the University of Leeds are using cutting edge technology to feed new lines to everyone’s favourite Italian-American goofball, Joey Tribbiani.
“I like pizza with cheese”?
“I love his girlfriend”?
It’s that old F.R.I.E.N.D.S magic alright, but what’s going on, and why?
What we’re looking at here is the ‘proof of concept’ (a POC is essentially a demonstration that proves the feasibility of a given idea, without demonstrating the full extent of the capability) of a research project titled ‘Virtual Immortality: Reanimating characters from TV shows’. As stated by the paper’s abstract, the aim of said project is as follows;
The objective of this work is to build virtual talking avatars of characters fully automatically from TV shows. From this unconstrained data, we show how to capture a character’s style of speech, visual appearance and language in an effort to construct an interactive avatar of the person and effectively immortalize them in a computational model.
As reported by The Verge the system is the result of de-constructing all 236 episodes of F.R.I.E.N.D.S and using the resulting data to develop new lines of speech that would be realistic in the context of Joey’s character.
The technology is of course still in its infancy and as such is a little rough and ready, but one of the interesting things as described by the paper is the relative ease with which the demonstration was achieved. The researchers were able to automatically analyze Joey’s facial movements and speech pattern throughout the entire series by manually working with only one episode, apparently only three hours’ work.
All well and good then, but things get very interesting indeed when considering the researchers’ predictions for future applications of their system, which include everything from developing visual representations of Apple’s famous sidekick Siri to creating interactive digital avatars of real people, potentially making dynamic conversations with real people, living and dead, a mind-blowing possibility.
I guess you could say that no-one told you that life was gonna be this way…