Hacker News Reader

Comment by kevingadd | Hacker News Reader

"AI will enhance you" is a really hard sell for voice acting. Voice actors are paid to use their physical training and experience in order to perform as directed in a unique way. Even if a voice actor were to go "now I trained this model on my own voice so I don't have to perform anymore", no sensible game studio or animation director would want to pay full price for an AI performance. The AI isn't going to be able to satisfy the precise demands of the director at a recording session, at least without a lot of domain specific training and extra work.

It's like telling a guitarist that AI will empower them because now the computer will play music for them.

An actor could probably additionally sell access to an AI version of their voice on the side, but doing that is devaluing their voice, arguably.

View on HN

> It's like telling a guitarist that AI will empower them because now the computer will play music for them.

I don't need AI for that - we've had drum machines for ages and now VSTs that can emulate pretty much any instrument. They do empower me because now I can create music that isn't limited to the instruments I own and can play without needing to hire an orchestra.

That's not empowering a guitarist to play guitar in some new or better or more efficient way.

Drum machines and VSTs are tools for any composer to synthesize a composition. They solve a different problem for a different set of people that can include but is not limited to guitarists. And as you pointed out they predate AI.

How do drum machines and VSTs make you better at playing guitar, or allow you to do that more efficiently or for higher pay? How will AI do that?

Don't guitarists record themselves playing a tune in a hundred different ways before settling on one that is the most appropriate?

> no sensible game studio or animation director would want to pay full price for an AI performance. The AI isn't going to be able to satisfy the precise demands of the director at a recording session, at least without a lot of domain specific training and extra work.

You’re assuming that it will be used for production. With a model than can spit out sentences on demand, game writers can try out a lot of different dialogue for draft storylines then have the final version recorded by the voice actor.

I'd 100% expect game studios to move to pure AI voices as soon as it's feasible to do so. There are even very strong use cases for it, like using it to voice generated dialogue so that NPCs don't need to be scripted word-for-word.

The entire genre of Internet video, too, is likely to do it, albeit because of cost.

Movies and TV may keep using voice actors if it's the prestigious thing to do. Who knows. The point about directing the actor is also good, but I don't know how long it would last.

TTS is already used for placeholder dialogue in games, yeah. That has been the case for a long time. There's not really a precedent for using fake versions of an actor's voice, but placeholder TTS dialogue has accidentally slipped into large releases before, and a few notable games have come out using AI instead of voice actors (The Finals, one of Cyan's games, etc)

Reply on news.ycombinator.com