The Evolving Face of Cinema: Exploring AI’s Influence on Movie-Making
Artificial Intelligence (AI) is radically altering the landscape of cinema production, from visual dubbing to the “resurrection” of past actors on screen, presenting a host of new possibilities and a few drawbacks in its wake.
In recent weeks, chatter has centered around ChatGpt, OpenAI’s new AI software, a San Francisco-based company backed by Elon Musk and Peter Thiel. Launched on November 30, downloaded for free, and already utilized by millions worldwide, it stands as the first mass AI bot. Its applications are limitless, capable of diagnostics, answering geopolitical queries, drafting letters, poetry, essays, and even informative code. And in case you’re wondering, yes, it can craft newspaper articles, just not this one.
The AI topic, following the momentary implosion of the metaverse bubble, seems to have revived discussions about the future and technology, penetrating even the most self-referential realms, like that of art. Towards the end of August, a work entirely generated by a program—Jason Allen’s Theatre d’Opera Spatial—created waves by winning an American photography contest. Alongside the expected outcry from human competitors, social media saw its fair share of “where are we heading, my dear” keyboard warriors. Admittedly, it triggered a common Pavlovian reflex. However, it pales in comparison to the shock of 1996 when Deep Blue defeated Garry Kasparov in a chess match. Since then, while chess players have evolved to become almost unbeatable, computers have progressed threefold more. The evolutionary path of machines, however, slipped quietly into acceptance, overshadowed by various emergencies, fleeting techno-media buzz, and clamor. Until ChatGpt emerged. Nevertheless, its applications continue to revolutionize various human activities, from agriculture to medicine, eliciting waves of expert enthusiasm and socio-anthropological alarm.
Cinema itself seemed to have relegated AI matters to the realm of imagination (Cronenberg’s ‘Crimes of the Future,’ the latest work, is more niche, an auteur’s dissertation). However, what exits the storytelling door can always re-enter through the production window. Towards the end of October, quite discreetly, our cinemas premiered an action movie, ‘Fall’ by Scott Mann. Centering on the misfortunes of two young women trapped two thousand feet above an abandoned radio tower in the desert, it was marketed to audiences as “an adrenaline-pumping experience exclusively for cinemas.” Yet, the film boasted a more exclusive credential: it was the first film extensively revamped by Flawless, an artificial intelligence startup capable of updating screen actors’ visemes (facial expressions) based on phonemes (the sounds actors produce). It was a substantial correction in film dubbing, defined last year by Time as one of the best inventions of 2021.
During the filming of ‘Fall,’ the two lead actresses, Virginia Gardner and Grace Fulton, were asked to improvise dialogues while suspended in mid-air. The outcome was so authentic that over 35 “f**k”s were counted from the young performers. Too many to evade the strict prohibition set by the inflexible MPAA, the American film authority comparable to our old censorship commission. Lionsgate, the production company, fearing profit erosion, urged the director to cut almost all profanities to attain a more PG-13 rating. There was only one problem: redoing those scenes, under those conditions, post-production, would entail considerable costs. It would require recalling the actresses, relocating them to the tower, reorganizing the set, and assembling the crew—a significant hassle. Enter Flawless AI: the software acquires the new dialogues recorded in the studio by the actresses and adjusts their facial movements to match the phonetics of the new dialogues. The result was astounding.
If applied on a broader scale, this technology could circumvent many translation and dubbing issues. Imagine the challenge of rewriting dialogues in another language while maintaining their meaning and synchronizing them with actors’ facial expressions. Flawless obviates any awkwardness, adapting not the dialogues but the mouth movements and even expressions for the most faithful translation. Americans call it “vubbing” or visual dubbing.
This isn’t the only instance of AI application in the film world. Concerning vocal-facial manipulation, Papercup is a company capable of generating synthetic human voices for use in dubbing and voice-overs. Here, the application is even more extreme: theoretically, subtitles and dubbing could be dispensed with because the software aims to make the same video available in any language. And what about Digital Domain, a visual effects giant, which has long been specializing in machine learning the images of actors when they are required to perform tasks that typically necessitate a stand-in? Then there are the extreme cases of deceased actors “resurrected” by AI in cinema engineering (think Peter Cushing and Carrie Fisher in recent ‘Star Wars’ sequels).
According to computer ethics expert Robert Wahl, AI will do to cinema what Photoshop did to photography, including ontological implications: can we truly believe what we see? If the philosophical question intersects with the more general concern about digital cinema and the blurring of the line between the apparent and the referent (think Pandora from ‘Avatar,’ existing solely in the film’s universe), the moral dilemma is more specific and pervading. The possibility of post-editing dialogues, attributing lines to actors for which they weren’t contracted, can become a riddle when the same actors are no longer able to accept or reject this alteration (perhaps because they have passed away). Not to mention the dangers, as highlighted by deep fakes, of false statements coming from real powerful individuals (or vice versa, false denials).
The proliferation of falsehoods or, if preferred, the foreclosure of truth, is the Pandora’s box (not the one from ‘Avatar’ this time) of AI applied to visuals. Despite the rightful assertion that technology is neutral, beneficial or malevolent based on its use, it inherently poses a problem. Quite prosaically, the American film industry is already collaborating with various trade unions to reach agreements on the conditions of utilizing these new computer technologies: ranging from consent policies (using them only if all parties agree) to financial compensations for professional categories that might be affected by AI implementation.
In summary, the American approach is typically pragmatic: business is business. No regulatory brakes: the train, after all, has already left the station. Who knows if it’s just an optical illusion, like that of the Lumiere brothers?
