Amid a lengthy contract negotiation with the video game industry, SAG-AFTRA has inked a deal with Replica Studios regarding the use of AI digital voice replicas in video games. The guild’s Na…
There are cool use cases for this such as having NPCs say player-picked names for their character (instead of saying a generic name/title like in current RPGs). I don’t know if it’s worth it though.
Are you suggesting replacing voice actors with a completely computer generated voice, for all voice lines? That hasn’t sounded good ever, especially in terms of intonation and emotions.
We are talking about a completely computer-generated voice. But for some reason we’re only talking about cases where it’s modeled as closely as possible on one specific human being.
Sure, and if it purely takes text as input and you use it to make voices for an entire game it’s going to be bad too. I’m talking about supplementing voices with snippets that they can’t know in advance, such as the player name.
I’m talking about artists using these tools for how characters sound, the same way they use tools for how characters look and move. The answer to “who plays Mario onscreen?” is “nobody.” The answer to “who voices Mario?” can be the same. It doesn’t mean nobody is responsible for how he sounds, or that someone typed “short plumber red hat” to render him.
There are cool use cases for this such as having NPCs say player-picked names for their character (instead of saying a generic name/title like in current RPGs). I don’t know if it’s worth it though.
Again: why would the NPC saying that need to sound like a specific real person?
To make it sound like the rest of the character’s dialogue. They’re probably not going to train this on the VA’s “normal” voice, that sounds useless.
Again: why would an NPC’s dialog need to sound like any specific real person, doing a voice?
Are you suggesting replacing voice actors with a completely computer generated voice, for all voice lines? That hasn’t sounded good ever, especially in terms of intonation and emotions.
That’s what this already does.
We are talking about a completely computer-generated voice. But for some reason we’re only talking about cases where it’s modeled as closely as possible on one specific human being.
Sure, and if it purely takes text as input and you use it to make voices for an entire game it’s going to be bad too. I’m talking about supplementing voices with snippets that they can’t know in advance, such as the player name.
I’m talking about artists using these tools for how characters sound, the same way they use tools for how characters look and move. The answer to “who plays Mario onscreen?” is “nobody.” The answer to “who voices Mario?” can be the same. It doesn’t mean nobody is responsible for how he sounds, or that someone typed “short plumber red hat” to render him.