Val Kilmer, the actor who played villain Iceman in 1986’s classic action film Top Gun lost his voice around 2014 during throat cancer treatment. Since then Vilmer cannot speak well, but thanks to an AI program, the audience was still able to hear Vilmer’s voice again this May 27 when Top Gun: Maverick, the sequel to Top Gun hit the theatres. 

Since its release, Tom Cruise’s Top Gun: Maverick promised to be a big deal. Although sequels aren’t always successful (especially after a couple of decades), the new Top Gun seems to have hit the mark. It’s making records at the box office, it’s being praised by critics, and it’s exciting for both fans of the first film as well as newer audiences. But there’s something else that’s interesting about this movie: Val Kilmer’s voice.

Recently, in an interview with Page Six, Val Kilmer’s daughter Mercedes Kilmer revealed that a special voice AI program was used to create her father’s voice in the film.

“They were able to dub him with his own voice, which is amazing. It’s such a technical feat, being able to engineer his voice that way, that it’s an extension of the technical feat of the film,” said Mercedes 

How did Kilmer get his voice back using AI?

Val Kilmer was the main antagonist in the first Top Gun film and therefore the filmmakers knew that the second part of the movie couldn’t be imagined without bringing in the Iceman (his character’s name in the film). In an interview with The Hollywood Reporter, Jerry Bruckheimer highlights that even lead actor Tom Cruise who also co-produced the film didn’t want to make the movie without Kilmer. 

However, there was one big problem – Val Kilmer couldn’t speak anymore as he had lost his voice during the tracheostomy and chemotherapy sessions he underwent because of throat cancer.

“I can’t speak without plugging this hole (talking about his throat). You have to make the choice to breathe or to eat,” said Vilmer in his biographical documentary Val.  

So the filmmakers decided to produce Kilmer’s voice for the film using archival footage and an AI-based voice dubbing technique developed by Sonantic, an AI technology company based in London. The team at Sonantic first collected Kilmer’s past voice recordings that didn’t contain any noise and then further cleaned the voice samples to produce crystal-clear samples of Kilmer’s voice. 

Transcripts of the clean audio samples were created and the audio was divided into smaller segments and paired with their respective transcript pieces. The team created an AI model of the voice samples and then trained the model with a voice engine using the smaller voice chunks and attached transcripts, essentially “teaching” the AI to follow Kilmer’s voice for different text inputs.

“With that limited amount of data, the algorithms we had been employing in our Voice Engine did not deliver the results we needed. We decided to research and implement new algorithms that could produce higher-quality options from the data we had available. In the end, we generated more than 40 different voice models and selected the best, highest-quality, most expressive one.”

AI is helping Hollywood with synthetic voice production

Interestingly, this is not the first time when AI was being employed to create an artificial voice for a real person. Film production companies and dubbing studios have been using AI models for enhancing sound quality, producing voices for different age versions of the same actor in films and TV shows, and recreating voices of dead personalities. For example, the 2021 documentary Roadrunner features a 45-second AI-generated voice of celebrity chef Anthony Bourdain who passed away in 2018. 

Moreover, the team behind the Star Wars spinoff series The Mandalorian also used 40-year-old sound recordings of actor Mark Hamill and AI-based synthetic voice techniques to create the voice for the young version of Luke Skywalker.

These recent developments strongly suggest that artificial intelligence is going to play a big role in shaping the voice dubbing industry in the future. Who knows, pretty soon we may not even need actors’ voices anymore — we could produce that synthetically. Whether or not that’s a good thing, however, is anyone’s guess for now.