Would you object the idea of CBU3 using Ai voice to voice player character?

[LordMcMutton]

I don't see support for the claim that generative AI entails theft. Not even Andreessen Horowitz makes the claim that AI development is impossible without free access to copyrighted material.

"Andreessen Horowitz is warning that billions of dollars in AI investments could be worth a lot less if companies developing the technology are forced to pay for the copyrighted data that makes it work."

I'm not sure what else you could glean from the fact that they're using copyrighted material without paying for it. (That's theft)

That's also just the only thing I happened to have on hand- I'm tired of wasting my time searching up sources for people that won't listen.

Other evidence is the fact that these systems can replicate specific people's art styles- the way they work means that can only be done by feeding them files of that specific person's artwork, which is using their work in a way they do not permit, which. Is. Theft.

https://www.fairlytrained.org/blog/n...odels-march-24

If it can be done, that's fine. What we have now is absolutely not.

You need to learn how to communicate without insulting others

I most certainly do not when my opponents' position entails support for mass theft and the destruction of my discipline.

In situations where it's merely a difference of opinion on something benign, then I do

You keep yelling that learning from others like a human is incorrect, then provide no evidence on why, other than a tweet by some artists who's public against AI. It's a fact that humans learn through trial and error iteration, just like AI. It's also a fact that human art students are taught to mimic other people's styles repeatedly, then they are taught to fuse styles together.

And the only evidence you have to say that they do is your own word, and the word of a bunch of idiots that zealously support "ai" garbage, so I'm not sure what point you think you're making here.

Additionally, I'd say they're far more qualified to speak on the matter than you are.

You say I have no idea how any of this functions, I studied as an artist for years, I went into the games industry for years and then left because most of the industry is incredibly toxic and pays poorly. I have a pretty good understanding of the industry and tools. An AI sampling your style is not theft any more that me sampling someone's style. You're basically against it because, machine, That's your only argument.

And? The fact that you spent all of that time in the field and still don't understand how this works is an indictment of yourself, not an argument for generative "ai" art theft.

[blah blah blah] sound similar to anything else..... kinda exactly how generative AI works.

Nope. For the reasons that have already been explained to you.

The only issue that's really there with Gen AI, is people trying to pass off Gen AI work as their own and asking for money for it. That's a tricky territory. As in, when you use Gen AI to make yourself a bunch of different version of something, you then work from that, tweaking and adapting it manually from there. Well is that the AI's work or yours now? That's a genuine question and something that needs to be discussed in law and industry.

It's always the system's work- you just 'commissioned' it.

No matter what additions you make, it's built off of stolen work.

End of story.

[Raven2014]

Even if you don't want to admit the "stealing" angle of AI - because granted, not all generative AI is trained on open/free/stolen works. I think most people are only aware of the general public AI like chatGPT, but we now have tools to let individual company to train their own AI.

But it's still a massive risk to employment. For example:

- A company hire an artist.
- Load them up with projects.
- Any work/art produced by the artist under contracts naturally belong to the company.
- The company uses these works to train their internal AIs.
- Once the AI is sufficient trained, fire the artist on the claim of redundancy.
- Now the company can continue generating almost identical work to the real artist without the artist themselves.

Legal? Perfectly. Moral? If you don't think this is exploitive, you're a fool. We already have phenomenom where employers fires veteran works to replace them with new/cheaper worker while forcing the veterans to train the very people who gonna replace them in the last months of the job. This will be even far worse. Eventually you will have agency that used to manage a pool of artists for commission, now will be just managing an AI instead.

Now some people may think that's not a problem, in fact I had heard some advocacy in the Music/movie industry that favor just pay the artists a lump sump contract upfront instead of the royalties system, and they may think what I said above is similar to that. Except it isn't. Without AI involved, while the company can continue benefit from the work they paid the artists for fair and square, they can not extend that benefit to competition, but AI can. So say company A hire the artist, trained the AI, then fire them. In the past maybe that's not a big deal, the artist can just go to company B. But now, with an AI trained, there is nothing to stop company A to compete with the artist for Company B's contract ... and probably win because the AI can do it much cheaper.

So yes, it's not necessary thief, but thieving is not even the biggest concern. Exploitive employers are very real, long before AI even is a concept, and it just gonna get a lot worse now. Exploit exist because they are legal, but a lot of them are simply just "legally stealing".


Edit to add: AI is here and it won't be stopped, that much is true. But that's why the real question these days is not really about how to stop AI, but to have regulations to protect real employees. Eventually I think we will see something like real artist gonna start patterning their art style, or employee demands in their contract to have clauses saying the company either can not use the work under contract to be used to train AI, or if it does they're entitled to something similar to a royalty. But these regulations gonna take time, labour law gonna need to catch up, and industry lobbylist will no doubt fight it at every bit on the way. I just hope they won't be too late before massive damage is done to the workforce. As a consumer I do consider it a moral obligation to try help - not to stop (because it's impossible) - but to buy time for new AI regulations to catch up. And I accomplish that by reject any company that actively using it to replace human.

[Iscah]

Also having an AI voice your character is not really the best route in an MMO. What would be better is having the main NPC's in cutscenes AI voiced. We're all wearing headsets with Mics or standing mics, so there is no reason you could not physically speak and have the AI NPC's interact with your actual statements. AI in the industry is early days but it's coming and coming fast.

I'm not here to play a game of "speak with the fake person to have a simulated conversation".

Invent new games with the technology for people who want to engage with a story that way, sure. But it isn't an inherent improvement on this game's approach.

[KisaiTenshi]

Why, Every silly argument that can be made to try and "save" the VO's has been used before in the past against media you currently happily consume, years ago people would make the argument, that video games shouldn't exist because they steal audience from authors.

What does that have anything to do with the subject.

Do I want AI voices in my video games? No. We have the evidence that audiences will not accept it from existing ways "AI characters" repeat the same lines and grunts over and over again.

FFXIV barely has any player-induced voiced sounds other than laughs and groans when they fall. It would get seriously annoying if the characters sounds one way until it has to say your name and you get "microsoft sam" every time. When it comes to permissive use of generative AI, it should be used as little as necessary. Generating sentences where the player's name is said, and localizations into languages that the game has not translated or dubbed yet are about the extent of generative AI's usefulness here. Generating dialog though? Forget that, LLM's could not write something interesting and funny because they don't understand the language. They are a parrot.

Generative AI can "translate" one voice into another language without needing a new voice actor. That Voice actor would still be providing the voice for the non-native language when the only alternative is "nothing" translated at all. The Voice actor still gets paid as though they dubbed those lines, and likewise for variable speech lines where the player's name is said, they're paid for that additional use as long as it's generated on the game server and not the game client (Which opens a pandora's box of players being able to make the AI voice say anything.)

Voice actors should oppose any dubbing of a video game with their voice unless they are paid for those lines as though they were actually performing them. That's the only way AI voices will be accepted by voice actors. Players however will never accept generative AI voices if they have to hear them all the time.

Shills for generative AI keep trying to push this idea that it can replace people in a creative sense, but that is never true, and will never be true. The AI has no experience, no wisdom, no third eye, no sense of embarrassment, shame, anger, happiness, love, etc. All it does is mash words together that fit together in LLM's and does the same thing with visual and audio signals.

[Emitans]

I have no interest in listening to an AI pronounce my name wrong.

[PyurBlue]

If it can be done, that's fine. What we have now is absolutely not.

It looks like this was a misunderstanding then. When I'm referring to AI I'm referring to the technology itself and not specific models. AI generated content does not necessarily have to come from AI that exist right now, and from a practical standpoint if SE were to incorporate AI into FF14 they would probably want one developed for the intended purpose if possible.

From the fairness aspect, if the existing ones are unfairly using training data then they should be excluded from consideration. That doesn't mean all AI should be excluded though. I'm happy to see that people are aware of AI issues and the problems that need to be solved, but totally opposing AI is not in anyone's interest, or at least I don't see how it could be. Let's try to solve the problems and make it work for us.

What does that have anything to do with the subject.

Do I want AI voices in my video games? No. We have the evidence that audiences will not accept it from existing ways "AI characters" repeat the same lines and grunts over and over again.

FFXIV barely has any player-induced voiced sounds other than laughs and groans when they fall. It would get seriously annoying if the characters sounds one way until it has to say your name and you get "microsoft sam" every time. When it comes to permissive use of generative AI, it should be used as little as necessary. Generating sentences where the player's name is said, and localizations into languages that the game has not translated or dubbed yet are about the extent of generative AI's usefulness here. Generating dialog though? Forget that, LLM's could not write something interesting and funny because they don't understand the language. They are a parrot.

Players should not settle for a minimum budget implementation. However I don't think you're giving AI enough credit. Things generally get better with time so AI voices are very likely to follow this trend. I can understand thinking that AI voices aren't ready for games today, but that will only last so long. AI's lack of true understanding also doesn't have to prevent it from being interesting or funny. The ability to predict language at all already goes a long way to aiding in those areas. If an AI can predict a typical and expected response to a statement then it can also find an unexpected one. This is an element of comdey.

https://improveconversation.com/2018...-and-surprise/

If you can create a set of rules for something then AI can probably be good at it, given enough time and understanding from the people developing the AI.

[KisaiTenshi]

Players should not settle for a minimum budget implementation. However I don't think you're giving AI enough credit. Things generally get better with time so AI voices are very likely to follow this trend. I can understand thinking that AI voices aren't ready for games today, but that will only last so long. AI's lack of true understanding also doesn't have to prevent it from being interesting or funny. The ability to predict language at all already goes a long way to aiding in those areas. If an AI can predict a typical and expected response to a statement then it can also find an unexpected one. This is an element of comdey.

https://improveconversation.com/2018...-and-surprise/

If you can create a set of rules for something then AI can probably be good at it, given enough time and understanding from the people developing the AI.

A lot of the stuff "Neuro-sama" does, that people like, is by accident. Accidental comedy by a LLM is no better than cherry picking madlibs. Here's the template to the "punchline", pick words that fit.

The reason you don't see a lot more "AI characters"/streamers that are actually driven by a LLM and TTS systems, is because people are only impressed by a parrot once. All these promises about what "AI might do" are moonshots. We are so extremely far away from anything that can mimic a human, and not also be tripped up by basic questions and behaviors, the people who think these silly parlor trick LLM AI's are sentient, should be kept far away from them, less they be convinced by one to harm themselves.

Which again going back to the question of using AI voices in a video game. People have tried this, human players of games reject them. There is a very narrow space where AI's are acceptable because they're not intended to replace a human, but rather as an assistive technology. For example ASR and TTS AI's that turn audio into text, and text into audio, don't require "Stolen" voices and text to be developed. However they do require more than "5 seconds" to make something convincing, and presently correct emotive responses are beyond any TTS's capability because all TTS systems are trained on narrative public domain speech, not casual speech. ASR's are trained against datasets that have been reviewed to be correct, but are woefully unreliable because it can't understand everyone's accents, and has no idea when something is filler words, onomatopoeia or slang. This creates some amazingly dangerous results like profanity being "heard" when opening soda cans. Ever watch youtube or twitch streamers with the automatic captions? The word error rate is so poor that it's barely usable. The ASR can only deal with speech as though it's being read like a 4th grader asked to read from a text book.

As an assistive technology, TTS is okay. It tends to feel like the entire anime subs vs dubs argument, where some people don't like to read, and others don't like the "cheapened" dubs (how a Japanese production will have unique talent for every character, but an English localization will reuse one voice actor as much as possible to save money.) The AI potentially solves this by keeping the original voice performance intact, but now can speak your language with the original voice actor, if you want it. It will not be as good as a native speaker, but it offers a third option. In the context of a game, this opens up the possibility of having one voice actor basically "perform" 100 other languages that they don't actually speak, and the game developer can't offer a dub for. It also works around unreasonable censorship that would be introduced by localization due to the "not invented here" word borrowing rules in languages like Chinese, French and Russian. It doesn't matter if there is a preferred word in the language, if the loan word is used in all the other languages, it gets used in all of them.

And FFXIV has a lot invented language words that can't be replaced with some languages invented modern words.

But I still think that the only reasonable, current, use of AI voices would be to personalize the game experience, and like the parlor trick above with AI characters, people are only impressed by this once, and it has to count, otherwise saying "warrior of light" and "she/he" during voiced audio is not the worst compromise. Some NPC's say our names, but only when they're not voiced. Many players, especially streamers, have basically taken the view that "Anything not dubbed, isn't important" and thus can just skip it. This is the impression given already when a game isn't completely voiced. So having some option to "Voice everything" would be nice to have to give all players an option to understand the lore, not just the players who like to read.

[Brynne]

I'm not in favor of using AI unless you're compensating the voice actors fairly for the use of their voices to "train" the AI. However, when you think of how many different voices you'd need to not only give us options to pick from but also in all the languages that the game currently supports, I think it would still be expensive and complicated. I'd rather SE spent the money and effort on other things.

[Oizen]

Only if my character sounds like Microsoft Sam from the Windows XP voice pack

[LittleArrow]

An idea that crosses my mind when playing modded Skyrim anniversary edition.

For those who don't know, there is a mod that can Voice your Dragonborn with AI voice pack,
which enable your Dragonborn able to talk like an actual character that is part of the world, with AI voice pack option ranging from God of War Ragnarok's Kratos to Genshin Impact's Raiden Shogun

Able to hear your Dragonborn talk is rather... refreshing to say the least

Then it's make me wonder the feasibility having our playable character voice with an trained AI Voice, considering our WoL has a lot of "Fuz Ro D-oh" moment in cutscene, especially with dialogue choices present.

(For those don't know what Fuz Ro D-oh is, it's a Skyrim mod that allow Dragonborn to move their lip when "talking", like how our WoL "speak" without any voice)

If one day Dev decided to include a feature that our WoL can actually talk,
yet again consider the amount of voice option each playable race,
the most economic and time saving way would be train AIs to match the respective voice option of our WoL

While the topic of AI in general tend to cause some degree of moral panic,
and given current track record CBU3, it might make them start to getting complacent in voice acting department, which is crucial part of MSQ experience, which yet a crucial part of entire FFXIV experience.

So, what do you guys think?

I wouldn't want AI or a voice actor to voice the lines of my character. This is pretty cringey, but I actually respond verbally to a lot of the npc's lol. Not seriously, but I often crack myself up with stupid silly responses that I think she would say. Also, I grown to enjoy that my character is the silent protagonist, so much that even the NPC's tease my WoL about it.