The Witcher voice actor Doug Cockle has expressed caution and frustration at the growing presence of artificial intelligence within the video game industry, calling it "inevitable" but "dangerous".
Voice AI isn’t going to replace all voice actors, and it won’t clone many specific people. Those are boring uses.
Voice AI is going to let a voice actor sound like anyone or anything… including people who do not exist. Because remember, these are fictional characters. Nobody plays Geralt, onscreen, in the Witcher games. Or if you like: Geralt is roughly modeled on some guy’s appearance, and acted out by a multitude of dudes covered in ping-pong balls. He’s a Geralt gestalt.
And he could be even less, if artists modeled and animated him without any “performance capture.” Like how nobody plays a cartoon character, onscreen.
Voice AI allows artists to invent audio the same way they can invent video. There might be more artists involved. Each of them churning out a whole library of voiced text for no-name NPCs. Honestly that’d be a pretty great way to recognize a dev team: make each of them a background nobody inside the world they’ve created, with any ridiculous details they want to throw in.
There will be no need to schedule a recording booth and approve every line months in advance. They can half-ass a take into their headset mic, tweak the machine’s output vaguely resembling their voice, and call it good. If the writers ever want to change something later on… they don’t need that exact person, in order to redo the lines. The differences might be noticeable. But probably less so than any two people doing Donald Duck.
Damn. You just made me realize some of the scope of background interest - that’s the sort of thing that Skyrim, GTA V, and Cyberpunk 2077 like to claim about NPC’s existing in a living breathing world on their own schedules.
You ever think about all the lives of the individuals next to you on the freeway? How all their lives are as complicated as our own? We don’t have that in games right now because it’s so complex and needlessly complicated for so little return.
But man, imagine a Shadow of Mordor connected NPC world. Friend groups, rivals, coworkers, bar patrons, they all have individuality that could affect other elements of the social game.
And yeah, current tech isn’t exactly good enough to propose a… David Cage roguelike, or whatever. Or GTA: Dwarf Fortress. Having those systems running all the time would be ridiculous, and worse, boring. But offline LLMs are definitely good enough to backfill details for any characters you do interact with. Like how Shadow of Mordor opportunistically promotes generic NPCs to rivals. Someone you drive past once doesn’t need a whole backstory and social network. But if you run them over, the game can give them a whole family, made-up on the spot. That part’s pretty easy. Making it interesting is the trick - and even if motives and consequences are shallow dice rolls, a modern chatbot can dress them up convincingly.
Voice AI isn’t going to replace all voice actors, and it won’t clone many specific people. Those are boring uses.
Voice AI is going to let a voice actor sound like anyone or anything… including people who do not exist. Because remember, these are fictional characters. Nobody plays Geralt, onscreen, in the Witcher games. Or if you like: Geralt is roughly modeled on some guy’s appearance, and acted out by a multitude of dudes covered in ping-pong balls. He’s a Geralt gestalt.
And he could be even less, if artists modeled and animated him without any “performance capture.” Like how nobody plays a cartoon character, onscreen.
Voice AI allows artists to invent audio the same way they can invent video. There might be more artists involved. Each of them churning out a whole library of voiced text for no-name NPCs. Honestly that’d be a pretty great way to recognize a dev team: make each of them a background nobody inside the world they’ve created, with any ridiculous details they want to throw in.
There will be no need to schedule a recording booth and approve every line months in advance. They can half-ass a take into their headset mic, tweak the machine’s output vaguely resembling their voice, and call it good. If the writers ever want to change something later on… they don’t need that exact person, in order to redo the lines. The differences might be noticeable. But probably less so than any two people doing Donald Duck.
Damn. You just made me realize some of the scope of background interest - that’s the sort of thing that Skyrim, GTA V, and Cyberpunk 2077 like to claim about NPC’s existing in a living breathing world on their own schedules.
You ever think about all the lives of the individuals next to you on the freeway? How all their lives are as complicated as our own? We don’t have that in games right now because it’s so complex and needlessly complicated for so little return.
But man, imagine a Shadow of Mordor connected NPC world. Friend groups, rivals, coworkers, bar patrons, they all have individuality that could affect other elements of the social game.
“Sonder” is the two-dollar word.
And yeah, current tech isn’t exactly good enough to propose a… David Cage roguelike, or whatever. Or GTA: Dwarf Fortress. Having those systems running all the time would be ridiculous, and worse, boring. But offline LLMs are definitely good enough to backfill details for any characters you do interact with. Like how Shadow of Mordor opportunistically promotes generic NPCs to rivals. Someone you drive past once doesn’t need a whole backstory and social network. But if you run them over, the game can give them a whole family, made-up on the spot. That part’s pretty easy. Making it interesting is the trick - and even if motives and consequences are shallow dice rolls, a modern chatbot can dress them up convincingly.