May 01, 2024 · Media

AI in media: who gets time back, and who loses the gig

Dimple Paratey
Dimple Paratey
Chief Marketing Officer
AI in media: who gets time back, and who loses the gig

There's a question I keep hearing at industry dinners, usually about three glasses in, from someone who edits or scores or illustrates for a living: "Am I about to be replaced?"

The truthful answer is annoyingly specific. Some of you, no. Some of you, the boring middle part of your job, yes. And a few of you who never got a foot in the door, the door just got smaller. Let me unpack that, because the hype merchants on one side and the doom merchants on the other are both selling you something simpler than reality.

Start with the wins, because they're real

A two-person studio can now pull off shots that needed a twenty-person VFX house in 2015. That's not work stolen from the big houses — the blockbusters still need them. It's work created for small studios that couldn't compete before.

Restoration is the same story. Peter Jackson's team used AI to clean up the Beatles footage in Get Back; Scorsese used AI-assisted de-aging in The Irishman. Painstaking craft that was impossible five years ago, now merely difficult. And the one that gets me every time: Val Kilmer performing in Top Gun: Maverick using a model of his own voice, trained from archival recordings after throat cancer took the real one. Built with his consent, by his own voice company. That is exactly the version of this technology that should exist.

Dubbing and subtitling have sped up enormously — human editors still do the final pass, but they cover more titles in more languages than they could before. And nearly every music producer I know is using stem separation, mastering assistance, smart sample browsing. They love these tools. As tools. Not as a replacement for the ear.

Now the part nobody at the dinner wants to say out loud

Library music is getting hollowed out. AI-generated beds are now good enough for podcast intros, corporate videos, game backgrounds — the bread-and-butter that paid junior composers while they learned. For the established names it's an irritation. For the juniors, it's the rung they were standing on, removed.

Voice acting for non-starring roles is going the same way. Cloning plus a competent director is, for a lot of commercial work, hard to tell from a hired actor. Audiobooks, ads, game NPCs. SAG-AFTRA didn't make this a bargaining red line for fun.

Concept illustrators who built a living producing mood-board variations are watching that work dry up. The sharpest of them are climbing into art direction or folding AI into their own process. Not everyone can, and pretending otherwise is a kindness to no one.

The pattern under all of it

Here's what I've concluded after a year of these conversations: AI is brilliant at average and terrible at idiosyncratic. Work with a distinct voice, a clear point of view, a reason only this person could have made it — that's getting more valuable, not less. What's being eaten is the competent-but-generic middle. The could-have-been-anyone work.

Which is the cruel bit, because the competent middle is precisely where most people learn their craft before they find their voice. We're automating the apprenticeship.

What I'd actually tell a studio or publisher

  • Lead with consent. If your model was trained on copyrighted work, sort that out before anything else — license the data or use models built on licensed data only. It's going to matter in court, and it matters before that.
  • Keep humans in the credits. If a writer, director, actor, or illustrator touched it, name them. Don't erase the labour that's still there.
  • Use AI to give your artists more time, not fewer of them. Strip the grunt work off skilled people so they do their best work more often.
  • Read what your union actually wrote. Those agreements aren't obstacles. They're the first rough maps of an industry that survives this.

Sound, colour, television, the internet — every wave was going to kill the industry, and every one ended up making more of it, eventually, after a painful stretch in the middle. I think AI lands there too. But "eventually" is carrying a lot of weight in that sentence, and the painful stretch is painful for people with rent to pay. The studios I respect are trying to make that stretch shorter and softer. The ones I don't are calling the layoffs "efficiency."

If you're a creative business trying to work out which side of that line you're on, that's a conversation we have a lot. Come and have it with us.

Dimple Paratey
Dimple Paratey
Chief Marketing Officer

Dimple leads marketing at Partech Systems. Before that she spent fifteen years in telecoms, mostly working in the gap between what the engineers built and what customers actually understood. She writes about the human side of technology — the people using it, the ones it tends to leave out, and the stories that get lost when we only talk in features and roadmaps.