And I addressed the expensive part. I’ll even argue that every playable character, even silent ones, are their own character. No game, besides Fallout 3, starts from birth and lets you be whomever as you go. But even in Fallout 3 you’ve still got many things pre-defined, parents, general backstory etc. Same with Skyrim, you’re the dragonborn no matter what you do. It’s really just a budget thing. To the point of the article though I think it adds to immersion to have a voiced protagonist. But others are of a diametrical opinion.
Sure, but none of those still lets you say whatever you want, you’re still locked to a few alternatives that effectively shoehorn you into a character. In the future, with the aid of AI you could match what you personally say or type to an intent and then map that to pre-baked responses. With attacking you or running away as a response to aggression and a “haha very funny but back to the matter at hand” when you’re being silly. And it’s only when we’re there that I’d, personally, say a “silent” protagonist is as immersive as a voiced one.
If I see a game with an AI voiced protag im avoiding it. Also I prefer my protags to be silent in RPGs, the constant yapping in fallout 4 pisses me off.
That tech or atleast the typing part of it has existed for a long time. Old text adventure games used it and even some of the earliest CRPGs used it like wasteland 1, also honorable mention to fallout 1 and 2 which had a watered down version. Theres a reason ya dont see it anymore, that reason is cause its fucking infuriating. Seriously half the time I try playing those I turn into a demented murder hobo.
I swear to fuck if I see a modern game go “I dont know what ‘talk’ is” im gonna punch the lead dev.
Yes and no, what they did is they used keywords and synonym lists but they were extremely basic and you needed to build a mental model of how dialogue “works” that fit the developers. Hardly ideal. But with LLMs and NLP and some clever programming making NPCs that react convincingly to your typed in questions / interactions isn’t far off.
They can’t handle a long conversation, that’s the problem. But no RPG I’ve ever played has very long conversations, especially not on one topic. The flow is generally opener -> pick a topic -> follow up questions (a few different that all land you back on the same “step”) -> conversation ender / go back to pick topic. Total sentences about one topic is generally no more than 10. And from the LLM standpoint it’d suffice if we kept the topic as one, joined, conversation and every distinct topic as their own instances. Thus we wouldn’t have time to dig to deep before the conversation ends and functionally resets.
But that, AI generated dynamic responses, wasn’t really what I meant. I more meant using player input in a dynamic fashion instead of selecting pre-baked sentences that your PC “says” without their own voice. So the flow would be that you typed or said “Hello there, can you tell me about the night the dragon attacked?” The AI would then interpret that text and match it to which Voice Actor recorded statement that best fit. So you’d get something like “Oh it was terrifying, the dragon burned down the church!” You’d get the same statement if you ask “whatever happened to that church over there?”
You’d also need recorded statements for rude remarks, aggression, that they don’t know (possibly with pointers towards NPC with more information) etc.
So you’re only replacing the dialog selector, not Voice Actors and you’re not voicing the protagonist.
And I addressed the expensive part. I’ll even argue that every playable character, even silent ones, are their own character. No game, besides Fallout 3, starts from birth and lets you be whomever as you go. But even in Fallout 3 you’ve still got many things pre-defined, parents, general backstory etc. Same with Skyrim, you’re the dragonborn no matter what you do. It’s really just a budget thing. To the point of the article though I think it adds to immersion to have a voiced protagonist. But others are of a diametrical opinion.
I was coming at this from a blank slate approach, think New Vegas, Morrowind, or Baldurs gate 3 ( assuming no premade character or Dark urge).
Sure, but none of those still lets you say whatever you want, you’re still locked to a few alternatives that effectively shoehorn you into a character. In the future, with the aid of AI you could match what you personally say or type to an intent and then map that to pre-baked responses. With attacking you or running away as a response to aggression and a “haha very funny but back to the matter at hand” when you’re being silly. And it’s only when we’re there that I’d, personally, say a “silent” protagonist is as immersive as a voiced one.
If I see a game with an AI voiced protag im avoiding it. Also I prefer my protags to be silent in RPGs, the constant yapping in fallout 4 pisses me off.
No I mean you’re the voice for the protagonist, the game reacts to what you say (or type if you prefer). And we’re all different in what we prefer!
That tech or atleast the typing part of it has existed for a long time. Old text adventure games used it and even some of the earliest CRPGs used it like wasteland 1, also honorable mention to fallout 1 and 2 which had a watered down version. Theres a reason ya dont see it anymore, that reason is cause its fucking infuriating. Seriously half the time I try playing those I turn into a demented murder hobo.
I swear to fuck if I see a modern game go “I dont know what ‘talk’ is” im gonna punch the lead dev.
Yes and no, what they did is they used keywords and synonym lists but they were extremely basic and you needed to build a mental model of how dialogue “works” that fit the developers. Hardly ideal. But with LLMs and NLP and some clever programming making NPCs that react convincingly to your typed in questions / interactions isn’t far off.
Given how schizophrenic LLMs can be im gonna guess it wont get past the experimental phase. Its just too damned unstable at the best of times.
They can’t handle a long conversation, that’s the problem. But no RPG I’ve ever played has very long conversations, especially not on one topic. The flow is generally opener -> pick a topic -> follow up questions (a few different that all land you back on the same “step”) -> conversation ender / go back to pick topic. Total sentences about one topic is generally no more than 10. And from the LLM standpoint it’d suffice if we kept the topic as one, joined, conversation and every distinct topic as their own instances. Thus we wouldn’t have time to dig to deep before the conversation ends and functionally resets.
But that, AI generated dynamic responses, wasn’t really what I meant. I more meant using player input in a dynamic fashion instead of selecting pre-baked sentences that your PC “says” without their own voice. So the flow would be that you typed or said “Hello there, can you tell me about the night the dragon attacked?” The AI would then interpret that text and match it to which Voice Actor recorded statement that best fit. So you’d get something like “Oh it was terrifying, the dragon burned down the church!” You’d get the same statement if you ask “whatever happened to that church over there?”
You’d also need recorded statements for rude remarks, aggression, that they don’t know (possibly with pointers towards NPC with more information) etc.
So you’re only replacing the dialog selector, not Voice Actors and you’re not voicing the protagonist.