Idea: Build an AI voice model for each player, and have an enemy use it to impersonate them.

[Krid]

We have all these AI models around, so why not have the game client build an AI voice model of the player, and then send over voice clips for an enemy to use to impersonate them?
It's a great way to use all the compute power the game doesn't need for graphics. :D

[Caveman]

No. No AI voices. I don't want a game cloning my voice, but I'd be fine if it works like one of the SCP games where it takes some of your voice clips, and replays them. Though in SCP, it's player controlled.
Maybe it would listen to phrases, and then repeats it.

[Sasha Mason]

Anything AI related is banned off of Steam, so they can't do that unless they want to start selling their game elsewhere.

[Isabelle]

this doesnt make any sense

you dont just "build ai voice models" in real-time using small snippets of low-quality user microphone data. these things require literally thousands of hours of high-quality voice audio.

there is no technology in the world, nor a strong enough computer to run it, that can even remotely achieve this

[Not Rawr]

Originally posted by Sasha Mason:

Anything AI related is banned off of Steam, so they can't do that unless they want to start selling their game elsewhere.

Thats just plain wrong homie.

This is interesting but sounds like a huge effort to implement and if there are any voice clips I'd prefer they be deleted after the session ends and not stored. I assume if they wanted to do this properly theyd need to collect a lot of voice samples over sessions.

[Bub]

Instead of using AI, the game could instead record everyone's mic and only use key words it catches to use as lures for players

Maybe the key words would be like "Hello", "Where are you", "Help me", "Are you there", "Where'd you go", "Can you hear me". Stuff like that instead of needing hundred of hours of audio

[Krid]

Originally posted by Isabelle:

you dont just "build ai voice models" in real-time using small snippets of low-quality user microphone data. these things require literally thousands of hours of high-quality voice audio.

there is no technology in the world, nor a strong enough computer to run it, that can even remotely achieve this

...rrrrright, which is why I never said real time. I also never said it should be reactive, or high quality. In fact, a partially trained model would be ideal because then it would *sound* off.
You seem to think I was asking for a perfect mimic capable of flawlessly impersonating the player and carrying on a conversation.
What I want is a jank mimic that hits uncanny valley if you're listening, but is somewhat passable if there's a lot of noise or you're not paying attention.

Originally posted by John Fortnite:

No. No AI voices. I don't want a game cloning my voice, but I'd be fine if it works like one of the SCP games where it takes some of your voice clips, and replays them. Though in SCP, it's player controlled.
Maybe it would listen to phrases, and then repeats it.

That's a fair argument.
It's why I said for the client to build the model and then send voice clips - if the model never leaves your own computer then it's not contributing to the growing dystopian nightmare we find ourselves in.

[Sleepy Sylvie]

Originally posted by John Fortnite:

No. No AI voices. I don't want a game cloning my voice, but I'd be fine if it works like one of the SCP games where it takes some of your voice clips, and replays them. Though in SCP, it's player controlled.
Maybe it would listen to phrases, and then repeats it.

definitely not, it's a major privacy concern. not to mention the legal issues of recording children and using them

[DJDiceZ]

Originally posted by Krid:

It's a great way to use all the compute power the game doesn't need for graphics. :D

Translation: I don't know how this works

[Coldflame]

how am i supposed to troll when this is the calibre of content being posted unironically

[The Fountaineer 🖋]

Sounds like a lawsuit just waiting to happen.

[Chaos]

Wow, you guys really want our newfound millionaire to lost everything through lawsuit eh?

[Axiolot]

Originally posted by Not Rawr:

Originally posted by Sasha Mason:

Anything AI related is banned off of Steam, so they can't do that unless they want to start selling their game elsewhere.

Thats just plain wrong homie.

This is interesting but sounds like a huge effort to implement and if there are any voice clips I'd prefer they be deleted after the session ends and not stored. I assume if they wanted to do this properly theyd need to collect a lot of voice samples over sessions.

that is literally the case lmao, valve pretty clearly said ai stuff isnt allowed on the store

[Sir Tontonolous]

Originally posted by Mrrrrp :3:

Originally posted by Not Rawr:

Thats just plain wrong homie.

This is interesting but sounds like a huge effort to implement and if there are any voice clips I'd prefer they be deleted after the session ends and not stored. I assume if they wanted to do this properly theyd need to collect a lot of voice samples over sessions.

that is literally the case lmao, valve pretty clearly said ai stuff isnt allowed on the store

https://www.videogameschronicle.com/news/valve-clarifies-it-doesnt-want-to-discourage-games-with-ai-generated-content/

https://techcrunch.com/2023/07/03/valve-responds-to-claims-it-has-banned-ai-generated-games-from-steam/

https://www.theverge.com/2023/7/1/23781339/valve-steam-ai-artwork-rejecting-banning-pc-games

Only if there's a risk of copyright problems cropping up, from everything I can see.

Check The Finals, that game uses ai voice models for its announcers and the game hasn't been booted off Steam or anything.

[Isabelle]

Originally posted by Krid:

Originally posted by Isabelle:

you dont just "build ai voice models" in real-time using small snippets of low-quality user microphone data. these things require literally thousands of hours of high-quality voice audio.

there is no technology in the world, nor a strong enough computer to run it, that can even remotely achieve this

...rrrrright, which is why I never said real time. I also never said it should be reactive, or high quality. In fact, a partially trained model would be ideal because then it would *sound* off.
You seem to think I was asking for a perfect mimic capable of flawlessly impersonating the player and carrying on a conversation.
What I want is a jank mimic that hits uncanny valley if you're listening, but is somewhat passable if there's a lot of noise or you're not paying attention.

Originally posted by John Fortnite:

No. No AI voices. I don't want a game cloning my voice, but I'd be fine if it works like one of the SCP games where it takes some of your voice clips, and replays them. Though in SCP, it's player controlled.
Maybe it would listen to phrases, and then repeats it.

That's a fair argument.
It's why I said for the client to build the model and then send voice clips - if the model never leaves your own computer then it's not contributing to the growing dystopian nightmare we find ourselves in.

You lack a very basic understanding of the things you are talking about. I literally cannot respond to this. This is complete nonsense

The resources required to do the things you are talking about would be ♥♥♥♥♥♥♥ astronomical. please educate yourself on the topic

[Razaah !!!!]

A monster that uses the last 10 seconds or so of voice from the player it killed like in SCP would be appreciated.
Those clips dont add to any model and will be deleted after the run on the moon ends.

A monster that has an approach like the ghosts of phasmophobia or other ghost games where the ai rcognized general phrases and answers according to them ( think spirit box)
would also be great fun. It could utilize the walkie talkies and be like a tell that some danger lurks in the facilities.