AI already does real work in the arts. It pulled one singer's voice off a scratchy tape, decades after he died. A camera rig made a 67-year-old actor look 25, in real time, on a working set. Translation software turned an actor's lines into convincing dialogue in another language, right inside the same shot.
Musicians, filmmakers and engineers use these tools now, on the record, this year. One question follows all of them. Who gets paid when a machine does the job a person used to bill for?
A dead man's voice, pulled off a cassette
John Lennon recorded a home demo in the late 1970s. He called it "Now and Then." He sang it alone, at a piano, into a cassette machine.
Paul McCartney and Ringo Starr tried finishing it in the 1990s. The piano on the tape bled into Lennon's voice. Only machine learning built decades later could pull the two apart cleanly.
Peter Jackson's team built that tool. They called it MAL, and built it while restoring old film for the 2021 documentary "Get Back." Jackson explained the trick in a 2023 interview. "During the course of Get Back we were paying a lot of attention to the technical restoration. That ultimately led us to develop a technology which allows us to take any soundtrack and split all the different components into separate tracks based on machine learning," he said1. The finished song reached the public in November 2023.
That splitting trick has a name, stem separation. A company called AudioShake now sells it to Universal Music Group, Warner Records and Disney Music Group2. It pulls solo vocal and instrument tracks out of old mixes that once held everything blended together.
Steve Greenberg runs S-Curve Records. He manages the band AJR, and he called the results a revelation. "To hear, in isolation, individual musicians' performances on tracks from the days before multi-track recording is a real revelation," he said2.
New faces on set, new words in the booth
Robert Zemeckis directed "Here." The film follows one family across sixty years, in one room. It reached theaters in November 2024. Tom Hanks plays his character at five different ages, and Robin Wright plays hers at four3.
A company called Metaphysic built the de-aging system. It ran live on set. Zemeckis watched one monitor show his real camera feed, next to a younger, digital version with about six frames of delay3.
Jo Plaete is Metaphysic's visual-effects supervisor. He credited the director's trust in new technology. "Working with forward-thinking filmmakers like Bob Zemeckis and Kevin Baillie was a privilege. Their faith in this novel technology pushed us to new heights and allowed us to deliver on their ambitious vision," Plaete said3.
Flawless AI builds a version of that software for voices, instead of faces. Scott Mann, a British director, built Flawless to avoid reshoots on his own film, "Fall," after a troubled shoot. Its tool adjusts an actor's lip movements to match dubbed dialogue. "About 80% of the workflow is automated," Mann said, "and it doesn't require any special equipment or direction during film production"4.
The same tool now works across Europe's dubbing industry, Reuters reported in 2025. Some voice actors there warn it could replace them. Mann argues the technology helps filmmakers when it is used well. "When AI technologies are used in the right way, they are a silver bullet to change how we can film-make in a new way," he said5.
Creators who got their voice back
Ask the people whose voices these tools rebuilt. The answer sounds almost grateful. AI gave some of them back a part of themselves that illness had taken.
Val Kilmer lost most of his speaking voice to throat cancer treatment in 2014. A company called Sonantic built him a new one, from old recordings. "Top Gun: Maverick" used that voice for Kilmer's one line in the 2022 film, with Kilmer's blessing6.
Patrick Darling is a musician diagnosed with ALS at 29. The disease took his ability to sing and play instruments. ElevenLabs, an AI company, helped him rebuild that voice from old phone videos and kitchen recordings, and in February 2026 he finally heard it. "The first time I heard the voice, I thought it was amazing. It sounded exactly like I had before, and you literally wouldn't be able to tell the difference," he said7.
Five named AI restorations reached the public from 2022 to 2026
Dated public releases of AI voice, de-aging and dubbing projects named in this piece
Each pin marks the date one project reached an audience, labeled by whose voice or face the tool rebuilt.
Sources: TVBEurope [1]; AWN (Animation World Network) [3]; Consequence [6]; MIT Technology Review [7]; CBS News [10]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Date | Event |
|---|---|
| May 2022 | Kilmer's voice rebuilt |
| Nov. 2023 | Lennon's voice finished |
| May 2024 | Travis sings again |
| Nov. 2024 | Hanks de-aged five ages |
| Feb. 2026 | Darling sings again |
Where the money actually goes
Executives see the same tools differently. CVL Economics surveyed 300 leaders across six entertainment industries in late 20238. Three in four said generative AI had already helped their company cut, shrink or combine jobs.
The same study looked ahead to 2026. It projected that 21.4 percent of film, television and animation jobs, about 118,500 positions, would be consolidated, replaced or eliminated by generative AI8.
CVL Economics expects AI to erase 21 percent of film and animation jobs by 2026
Share of U.S. film, television and animation jobs projected consolidated, replaced or eliminated by generative AI by 2026
Source: AWN (Animation World Network) [8]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Item | Value |
|---|---|
| Jobs projected gone | 21.4% |
| Jobs expected to continue | 78.6% |
| Whole | 100% |
Steve Kaplan negotiates contracts for the Animation Guild9. He told studio negotiators in 2024 that members wanted limits on the technology. "You told us that addressing the impact of generative AI tools and processes are of the greatest importance and that you want us to make sure jobs in animation are made for human beings and not software routines," Kaplan said.
Money moves strangely even in stories that look like clean wins. Warner Music Nashville pulled James Dupré's 2011 demo vocal from storage10. It became the base for Randy Travis's AI vocal on "Where That Came From," released in May 2024. Dupré had already been paid once, back in 2011, for a version of the song that stayed in the vault for over a decade.
"I really just thought he recorded it years ago and I never knew about it. Then they told me about the AI and I was just so fascinated," Dupré said11.
Warner confirmed Dupré's 2011 payment. It left open whether he would be paid again for the version built from his voice11. The song debuted at number 45 on Billboard's Country Airplay chart, Travis's first chart appearance there in nearly 20 years12.
What to do
Read any AI-in-the-arts story for who actually gets the time back. A tool that saves a studio two weeks of dubbing work hands that time to the studio first. The contract then decides whether the performer it replaced ever sees a share.
Ask one question before praising or criticizing a new AI credit. Who owned the original recording? Does the new contract pay the person whose voice, face or performance trained the tool? James Dupré's unresolved pay shows that question stays open, even inside a charting song11.
By the numbers
- 45 years: the gap between John Lennon recording his "Now and Then" demo and Peter Jackson's software finishing the track1.
- Three major labels, Universal Music Group, Warner Records and Disney Music Group, now buy stem separation from AudioShake2.
- Five ages: how many versions of Tom Hanks appear on screen in "Here," thanks to Metaphysic's de-aging system3.
- 80 percent: the share of Flawless AI's dubbing workflow the software itself completes, Scott Mann says4.
- 118,500 film, television and animation jobs are projected gone by 2026, consolidated, replaced or eliminated by generative AI, CVL Economics found8.
- 45th on the chart: where "Where That Came From" debuted on Billboard's Country Airplay chart, Randy Travis's first chart appearance there in nearly 20 years12.
What to watch
Watch whether credits start naming the voice or face behind an AI tool. This piece names Dupré; most AI projects still skip that credit. European voice actor groups want EU rules that require consent and pay before training on an old performance, Reuters reported5. A similar rule in Hollywood's next contract would answer the question this piece opened.
Sources
- TVBEurope, "Watch: Peter Jackson discusses the machine learning technology behind Now and Then," Nov. 3, 2023, https://www.tvbeurope.com/artificial-intelligence/watch-peter-jackson-discusses-the-machine-learning-technology-behind-now-and-then
- Daniel Tencer, "AI music startup AudioShake raises $2.7M in seed round with backing from Crush Music, peermusic... and Metallica," Music Business Worldwide, May 9, 2023, https://www.musicbusinessworldwide.com/ai-music-startup-audioshake-raises-2-7m-in-seed-round-with-backing-crush-music-and-peermusic/
- AWN, "Metaphysic De-Ages Robert Zemeckis' 'Here' Via Generative AI," Nov. 21, 2024, https://www.awn.com/news/metaphysic-de-ages-robert-zemeckis-here-generative-ai
- TNW, "Flawless AI's 'True Sync' tech is a revolution for film dubbing," Nov. 28, 2022, https://thenextweb.com/news/london-based-flawless-ais-true-sync-tech-is-a-revolutionary-approach-to-film-dubbing
- Elizabeth Pineau, Miranda Murray and Dawn Chmielewski, "Voice actors push back as AI threatens dubbing industry," Reuters, via RTE.ie, July 30, 2025, https://www.rte.ie/entertainment/2025/0730/1526154-voice-actors-push-back-as-ai-threatens-dubbing-industry/
- Consequence, "Top Gun: Maverick used A.I. to recreate Val Kilmer's voice," May 31, 2022, https://consequence.net/2022/05/val-kilmer-ai-top-gun-maverick/
- Jessica Hamzelou, "ALS stole this musician's voice. AI let him sing again.," MIT Technology Review, Feb. 13, 2026, https://www.technologyreview.com/2026/02/13/1132913/als-stole-this-musicians-voice-ai-sing/
- AWN, "The Animation Guild Shares 'Generative AI' Economic Impact Study Results," Feb. 12, 2024, https://www.awn.com/news/animation-guild-shares-generative-ai-economic-impact-study-results
- TheWrap, "'No Contract, No Cartoons': Animation Guild Members Stand Together as Talks Begin," Aug. 12, 2024, https://www.thewrap.com/animation-guild-members-contract-negotiations-amptp-iatse-rally/
- CBS News, "More than a decade after a stroke, Randy Travis sings again, courtesy of AI," May 5, 2024, https://www.cbsnews.com/news/randy-travis-sings-again-courtesy-of-ai-where-that-came-from/
- American Songwriter, "Who Is James Dupré? The Real Voice Behind The AI Randy Travis Song," May 9, 2024, https://americansongwriter.com/who-is-james-dupre-the-real-voice-behind-the-ai-randy-travis-song/
- 117 Entertainment Group, "Randy Travis Debuts On Billboard Country Airplay Chart For First Time In Two Decades," May 13, 2024, https://117group.com/randy-travis-debuts-on-billboard-country-airplay-chart-for-first-time-in-two-decades/