Voice Acting and AI

 It's the future, but not as you know it

Voice acting is one of the most integral aspects of Modern Game design that contributes towards elevating a gaming experience to truly immersive precipices. At it's basic most principle, Voice Acting allows the voice of the fictional world to be made real and breathes that touch of humanity into the game, in a manner that is tangible and to which people can form an emotional connection. That isn't to say this is the only manner in which emotion can be established, but it's one of the most accessible ways, thus the steady improvement of video game voice acting standards has coincided with an increase to more emotionally charged narratives everywhich way on the gaming spectrum. Voice acting is so important to modern video games that even Bethesda lost their minds one year and decided that the protagonist of Fallout 4 should have a voice actor, (Hopefully they're never going to make that mistake again.) purely because most other modern games capitulated to that industry standard and it felt like the thing to do. Thus when the voice acting community feels like it's future is being threatened by the march of game industry technology, it inspires a serious degree of introspection in deciding where exactly the issue is and if accords can be reached.

Recently some controversy sprang up around the completion and release of a much anticipated mod for the 'Witcher 3: Wild Hunt' called 'A Night to Remember'. (Inspired by the teaser trailer for the Witcher 3 of the same name) The Mod in question acted as an expansion upon the events of the Blood and Wine expansion and would follow the protagonist Geralt on a whole new adventure for wrapping up a hanging quest line. There was just one nagging problem; The Witcher 3 was a highly cinematic RPG where every single character was voiced with memorable gusto, so how could the team emulate that base game content without a voice? This has been an issue in modding stemming all the way back to the days of Oblivion, when voice acting first became a standard in The Elder Scrolls. Suddenly most modded content, usually developed by a single person in their spare time who doesn't have access to multiple voice actors, really stood out from the vanilla game through merit of it being voiceless.

Some modders have found ways around this, most amusingly through the recycling of other in-game audio to form new sentences, which is a practise that has become frighteningly effective over the years. Others just resort to seeking out amateur voice actors or even hiring professional ones, whatever it takes to make that project which took years to drum up stick the landing. But usually, when it comes to recycling lines or hiring an imitator voice actor to take over from some character who's already voiced, like the protagonist Geralt from The Witcher, results are either restrictive or unspectacular. Especially when we're talking about a performance as iconic as Doug Cockle's, how do you live up to a voice like that? Which is where a whole new controversy-raising solution came into the mix when this mod's creator managed to make use of some AI tech.

Essentially, with my limited understanding of technology, it would appear that some sort of software has allowed previous voice lines of Mr Cockle's to be fed into an algorithm that then takes the sounds, memorises them, and reshapes them into entirely new lines. The effect is... really good. As in, if I wasn't already told that this was the case I'd have just assumed that the developer reused some voice lines and maybe dug into some cut content for the rest. Now bear in mind two facts as I say that, that I've only seen a little bit of it in action and there's bound to be some seriously notable slip-up in the full performance, and that this is Geralt's voice we're talking about here, which is rather characteristically monotone. I'd imagine that with other characters that tend to shift their tones and be just a tad more emotive we might have a few hiccups in the system. Still though, at face value this is some incredible technology.

Which is where the concern comes into it, because when it comes to mimicking the voices of voice actors one has to wonder about the potential impact this could have on the industry if it were to become a widespread practise. I mean, take a look at what decent CG has done to the acting world, wherein famous deceased actors can now have themselves digitally imposed into movies to reprise their roles and take on new ones, irregardless of what the actor themselves might have thought. There's a question to be raised there of morality and ethical treatment, which doesn't quite touch this voice acting issue just yet, but there are other parallels that do. Take, for instance, the fact that there's an upcoming movie with James Dean in it. Yes, the dead one. Now I'm sure an actor will be under the CG, acting as James and maybe even providing his voice, but that actor won't be getting the exposure and recognition that he would were he starring in that movie himself. With Voice acting you don't even need the body underneath, developers could cut out the need for voice acting entirely. At least in theory.

In practise there's a lot more nuance to this issue that makes it not the 'looming death knell to voice acting as we know it' that it might seem. Not least of all potential legal ramifications of copying someone's voice, as highlighted by voice actor Jay Britton over Twitter. Obsidian recently revealed how their use of such technology, in it's rudimentary state, is simply to give voice to lines that they write so that they can hear how it sounds and decide what works and what doesn't. As an amateur writer myself I can tell you that is a huge part of deciding the soul of your dialogue, so I can absolutely understand how this technology can be useful, without it also being harmful to the acting industry. And that's not to touch on all the imperfections that, no matter how good this tech gets, it will theoretically never master. (Or, at least, not until that AI Event Horizon I'm talked about before)

As this issue jumped about on the Internet, several voice actors came out to explain the many ways in which their job cannot be taken by a computer due to the many talents that only a human mind can intentionally work with. Some talked about 'breathing' in-between and during lines, which is said to be a big part of affecting the heart and intention of dialogue. Others mentioned the experience that an actor brings to a role, citing the famous story of Christopher Lee informing Peter Jackson how one would sound after being stabbed in the back, drawing from his own history in World War 2. One took a more antagonist approach toward technology in general, painting a dystopian hypothetical wherein an actor's voice and their character becomes iconic through no work of the actor in question, concluding how it's devoid of 'life' and 'art'. Most valid points, however I will push back a little as this is very much Art, just art of a different kind. Software art. And if you consider that invalid through sheer merit of it being from a computer then you might just find yourself fighting against windmills someday.

But as for today the argument is moot, because there truly is a lot more value a human being can bring to a role than an AI possibly could. One company behind a branch of this AI software has cited that this sort of tech would be ideal for a game NPC who needs to take in new information and respond to it, but given the inherently limited nature of a constructed game, as well as the fact that lines still need to be written by an actual person for the moment, I'd say there's little no real world example for that right now. Still, it opens an interesting, and frightening, proposition to what the future might hold for voice acting across the world, because you can bet your bottom dollar that the second this sort of tech becomes good enough to replace real people EA and friends will drop on it like Flies to excrements. I only wonder if there's anything between now and then we can do to protect those who's jobs might be threatened.