What if a meeting with Dalida, Jean Gabin or Lady Di was possible in 2022? We would dream of it, Thierry Ardisson has (almost) done it. In Time Hotel, the first edition of which will be broadcast on France 3 on Monday 2 May, the star presenter interviews deceased actors, singers and politicians. How is it possible ? By using one of the applications of artificial intelligence: deepfakes. More precisely, the presenter and his team relied on the technologies of two French companies, Ircam Amplify and Mac Guff, to recreate the voices of the missing stars with one and their faces with the other.
Behind the algorithmic experiment, two French companies
“One of our areas of work is voice in the broadest sense,” explains Nicolas Pinngnelain, commercial director of Ircam Amplify. Founded in 2020, the company aims to economically promote the fundamental research conducted at the Institute for Acoustic/Music Research and Coordination (Ircam). “Upstream this means detecting emotions, markers of stress or anxiety, downstream, synthesizing a voice that can reproduce this kind of variation. Man is even more sensitive to the voice than to the image, he continues, a trifle sets us in able to detect whether it is false No wonder Thierry Ardisson’s team turned to French researchers: a monotonous voice like that of our intelligent assistants to take over Dalida’s intonations would never have worked.
Also for the faces it was necessary to create the illusion of reality. The team turned to Mac Guff, a 35-year-old French studio, maker of Me, ugly and mean† “We have always had a research and development dimension, which allows us to stay ahead of the curve,” said Chairman Rodolphe Chabrier. For example, we were one of the first to do morphing, relief…” Five years ago, the entrepreneur started with artificial intelligence. Mac Guff soon built two tools: the Talking Picture, to make fixed images “talk”, and the Face Engine, especially for recreating faces in 3D. Both based on deep learning models, these technologies made it possible to model the faces of stars of the past on the bodies of the actresses and actors who gave the answer to Thierry Ardisson during the filming of the docu-fiction.
How do we raise the dead?
Because Thierry Ardisson loves it: Time Hotel is documentary work. And it’s true: to train the algorithmic machines that reconstruct faces and voices, you have to equip them with equipment. These training data sets consisted of all possible records about each of the celebrities the show brings to life. “Television appearances, images of families, films in the case of actors like Jean Gabin, Rodolphe Chabrier sums up… We take everything we find! “For the voice, it’s a bit easier, explains Nicolas Pingnelain: “The most important thing is to have samples of all phonemes, all syllables that occur in the language. Afterwards, we can achieve a good result with a few tens of minutes of speaking time. “As for the words of Dalida, Lady Di or Coluche in the show, Thierry Ardisson underlines, they were all really said during the celebrities’ lives.
The result works quite well… But we still feel uncomfortable. Could it be the Valley of the Uncanny effect, which makes robots trying to imitate humans a little off-putting to the person looking at them? Is it the fear of a certain smoothness of the faces, when one lingers on the nose a little too fine of the artificial Dalida, in the photos of shootings? “We had Dalida’s voice validated by her brother Orlando just two weeks ago and he cannot tell the difference between the synthesized version and the real one,” assures Nicolas Pinnnelain. “When they saw the result, the Gabin sons found their father’s expressions, they were amazed! adds Rodolphe Chabrier. The production also made the decision to “enlarge” the stars, he specifies, “to look for the ‘iconic’ Dalida, as Thierry says”. In this case, the unease may stem from ethical questions: is it acceptable to bring dead people back to life? Response from everyone who contributed hotel of the time : Not only is there no law that prohibits it, but the rights holders have looked at everything, the scripts, the faces, the voices, and everything has been validated.
“A tool has no ideology, emphasizes Thierry Ardisson, this point is very important in the light of the a priori that may exist”. Rodolphe Chabrier conveys the same message and we can’t help but smile: in tech, the argument of neutrality is a classic. However, in the world of artificial intelligence, more and more researchers and activists believe that posing the problem in terms of good or bad use is not always relevant: in some cases, it would simply be necessary to remember to use a tool. to develop. For hotel of the timethat said, it’s clear that Thierry Ardisson follows an idea that has long obsessed him: In 1994, he “awakened” John Lennon in Gone with time† In 2002 it was Victor Hugo he invited to the set ofWe will have seen it all, alongside Renaud and Guy Bedos, alive and well. Replacing lookalikes with deepfakes, maybe it just changes with the times?
This would be to miss all the innovation dimension that the creation of the program required: Rodolphe Chabrier like Nicolas Pinnnelain underline the innovative aspect of the program and the work it required. “60 minutes of reconstituted faces over 90 minutes of program, it’s never been done before! exclaims Mac Guff’s leader, if only because the technologies accessible before the company built its own algorithmic models would have made the operation far too expensive. For voice synthesis, “our researchers modified the tool less than a month ago, and they will likely do it again after the release ofhotel of the time says Nicolas Pinnnelain: the show used applications straight from basic research. Any scientific advancement will improve the programs built for the occasion.
A double cultural opportunity
In fact, maybehotel of the time opens up two cultural axes at once: first the opportunity to get to know the interviewed stars better. After all, Thierry Ardisson’s stated goal is to “spectacularize culture and make knowledge fascinating”. It is also the chance to see what a still young technology can bring in the field of visual and sound special effects. Deep forgeries have already been used in series – by Mac Guff, to rejuvenate Mathieu Amalric in The office of legendsor by Russian operator Megafon, to play Bruce Willis in an ad without actually acting.
Difficult to predict whether hotel of the time will mark a turning point itself. On the other hand, working on this particular case enabled the Mac Guff developers and the Ircam researchers to make progress in their respective fields. And among the uses they’re coming up with for their new algorithmic tools, Rodolphe Chabrier and Nicolas Pinnnelain outline ways as varied as voice authentication (to finally give up passwords), dubbing (no more French voices that don’t match American actors?) or , more classically , retouching faces and then bodies. “The Face Engine allows you to recreate, rejuvenate, age, grow, transform into a burn victim, for the cinema or video games in a very realistic way,” illustrates Rodolphe Chabrier. Who knows what show Thierry Ardisson will invent when it becomes possible to play on bodies and environments as well.