AI-generated voice is good/bad?

[Frindis]

I was inspired to make the thread after reading @neogunhero post about voice acted character speaking robotically in this post: https://forums.pcgamer.com/threads/which-game-features-are-you-tired-of.139687/post-355454 . It made me think about a Twitch stream by Asmongold where he looks at a clip with a WoW addon that plays out an AI-generated voice for NPC quests in World of Warcraft after being generated through an AI tool called Evenlabs.

View: https://www.youtube.com/watch?v=Ppp3diO5O18


So, my question to the community is where you stand in all this. Do you think it is ok to have an AI-generated voice if it is better than a voiced character speaking robotically or would you always prefer it to be voiced by a human regardless of it being good or bad because anything is better than using AI?

My stance is to use it if it can bring good quality to the game and I also think this could be a great addition for indie developers who might want to save a few bucks instead of hiring expensive voice actors or making bad voices themselves because they are poor. I have played enough poorly voiced indie games to know just how bad it can be. I can see voice actors selling services for similar tools also, so I don't think it will be something that steals all the work for the voice actors.

On a side note Blizzard has recently made a patent about using more AI in their games, which has gotten criticisms because handcrafted has always been the staple for the franchise.

 

[mainer]

I can see voice actors selling services for similar tools also, so I don't think it will be something that steals all the work for the voice actors.

This is what I hope to see happen, or something similar, like a voice actor recording much of the VA for a character, then have the AI applied to that character to create the spontaneity of life-like conversations and reactions. The difficult part for the AI, will be getting the emotion and subtle inflections into the dialog. I think there's enough potential for both voice actors and AI to function together.

I would never want to see AI completely replace actual voice actors. Some of the RPGs I've played have had really excellent voice acting, like the Mass Effect series, and those actors are what really gave those games emotion that the player could feel while playing.

We're still in the infancy of AI generated voices, where people are experimenting, and most examples are basically prototypes with emotionless voices. But you have to start somewhere, and I do think the tech will advance quickly, though eventually I also think we're going to have to have some regulations in place so that everyone is happy with the results. Well, I guess everyone will never all be happy, but maybe finding a proper balance between human voice acting and AI.

 

[neogunhero]

As said above I’d hate to see AI fully replace traditional voice acting, but I agree with using AI to help enhance dialogue in games. My main gripe was that the flow of dialogue does not sound realistic or human in a lot of games, most notably the Assassin’s Creed series in my opinion. One person speaks, while the other waits their turn to speak. Realistic human conversation has a lot of interrupting, stammering, thinking out loud, etc. My theory on the cause of this is that the game developers has to program the game to handle these dialogue sequences, so perhaps it’s harder to make it flow more realistically? This does not apply to scripted scenes like on The Last Of Us games where there are fully acted cutscenes. I’m mainly talking about games like Skyrim, Assassins Creed, or Horizon Zero Dawn where you talk to hundreds of different characters and have options on what to say. Of course you may have to pause to make a decision on what to say next, some games do away with that by giving you a timer to choose but I find that much too distracting, having to think of my choice plus listen to what the other person is saying makes my brain no work good.

Perhaps AI can help take these pieces of recorded dialogue and weave them into a realistic flow of human-like conversation. Also said above, perhaps it can be trained based off a voice actors recorded lines to create variations of their lines and random shouted dialogue while in game.

I may be crazy as I haven’t done concrete investigation or seen any videos on this, but I swear on RDR2 if you’re having a conversation with someone while you’re both on horseback, they will start to shout at you if you get far from them so they can hear you better. If that’s the case that takes some serious time and effort to record all the different variations! AI could potentially help with this by taking a recorded voice line then creating different variations of it, such as the character talking louder towards you when you’re far away, or maybe start to whisper as your crouch and begin to sneak. It could also help with those random lines that NPCs yell at you as you walk past them. All of this could help boost immersion and realism.

We’re at the very beginning of seeing how AI can help with game development and it’s all extremely exciting. My main stance still stands though, I would hate to see AI fully replace anyone’s positions, in any industry or job. Let’s use AI as a tool to help human creativity, not to replace us.

 

[Frindis]

I may be crazy as I haven’t done concrete investigation or seen any videos on this, but I swear on RDR2 if you’re having a conversation with someone while you’re both on horseback, they will start to shout at you if you get far from them so they can hear you better.

I do believe you are right about that because I remember this being the case in the intro when you ride together with Dutch. I'm not sure if it is only in the scripted scenes or if also random NPCs would do it. I would not be surprised though, this game has everything from standing still and you might avoid a grizzly attack to the sound of a rifle shot echoing in the canyons to shrinking horse balls in snow. It almost seems like they have thought about everything in this game.

 

[Brian Boru]

this could be a great addition for indie developers

That's the main short-term benefit. The cost of voice and art assets are prohibitive for any indie effort which doesn't get some serious crowd funding, so cheap tech to allow people to test if their design and implementation skills are real or imaginary should lead to an explosion in creativity.

It's like printing. Previously only aristocracy who could employ scribes were able to get material into circulation.

I would never want to see AI completely replace actual voice actors

Shouldn't happen for a long time, if ever. The voices in the Far Cry games have always been a big part of their appeal, hard to see AI being able to offer equivalence to Giancarlo Esposito, Michael Mando, or Troy Baker anytime soon.

The huge plus for AI is what you guys mentioned, more variety in the random NPCs artwork and voices, plus dynamic dialog.

Realistic human conversation

Is that something we want repeated in games? Why bring those weaknesses you mention into games? I much prefer non-gibberish in my game characters… leave the realistic stuff to add color to cut scenes, if someone thinks it has a place.

I would hate to see AI fully replace anyone’s positions, in any industry or job

Why?

The history of almost every peaceful tech advance has been a big leap in human wellbeing—I mentioned printing, look at agriculture, replaceable parts, vaccines, electricity… it's a long list.

There is general agreement that we currently live in the best era for humanity, and there's every reason to expect that to continue. There are always transition pains to be sure—eg the industrial era to bridge the gap between artisan individuality and universal material goods has certainly been a bumpy couple of centuries, but should be over this century and the result be a huge boon to humanity from 22nd century on.

Unless humans screw it up

 

[Zloth]

It's big, that's for certain. My thoughts are scattered all over the place...

• Using AI for bit parts sounds great. Remember Elder Scrolls: Oblivion's use of just a few voice actors for so many characters? Gone. Give everyone a unique voice.
• Flexibility of lines will massively improve. Need to change the name of your kingdom because the name you have means something naughty in Italian? No problem. Even for the parts done by voice actors!
• You know how a very few games will let you pick between 4 or 6 voice actors for your main character? How about having sliders so you can design your character's voice with as much fine detail as you like?
• On the other hand, how good is the AI about reacting to other speakers' lines, or even their own past lines? If the code is just analyzing sentence by sentence, it's going to have some wild mood swings. Even analyzing entire paragraphs may not be enough.
• Writers are probably going to need to adjust. When they invent new words (or even just use rare ones?), they'll need to give the phonetics somewhere. They might need to help with the emotional parts, too. (Like writing for the Elcor in Mass Effect.)
• No way in hell are voice actors going to sit still while we lop their economic heads off. They've got a union and they're going to use it. It won't be isolated, either - we've already got a writer's strike over here. And then there's all the worry about the fake videos that's bound to get mixed in with it all. We'll be talking about this a lot - same as when the internet first got big.

One person speaks, while the other waits their turn to speak. Realistic human conversation has a lot of interrupting, stammering, thinking out loud, etc. My theory on the cause of this is that the game developers has to program the game to handle these dialogue sequences, so perhaps it’s harder to make it flow more realistically?

No, that's not just games. Movies and TV shows do the same thing. Plays going back centuries do, too. Realism is nice, but people just want to understand the story, so it gets cleaned up. (Besides, in today's vernacular, a lot of those pauses and stammers would turn into the world "like" splattered all over the sentences. Do you, like, really want that?)

P.S. We aren't at the beginning, folks. Text-to-speech has been around for decades. We're at the beginning of it not sounding awful.

P.P.S. I need to find a way to hack into a lot of training datasets and make sure "Kraken" uses the long 'a' sound.

 

[SammieAdams]

I hate AI, it makes us more lazy.

 

[ZedClampet]

So, my question to the community is where you stand in all this. Do you think it is ok to have an AI-generated voice if it is better than a voiced character speaking robotically or would you always prefer it to be voiced by a human regardless of it being good or bad because anything is better than using AI?

I refuse to be a pointlessly looking backward rube. Gimme the AI. AI can sound perfectly natural (see the Todd Howard "speech"). There's no point in not using it just because we're scared of the future.

Note: I didn't read the thread, so no offense was intended towards pointlessly looking backward rubes who may have posted before me.

 

[Brian Boru]

@ZedClampet The Goldberg family are a bit miffed at the unfortunate association, but on the other hand are using AI to program their next machine.

 

[Khushi100]

It's an interesting debate! In the end, the choice between AI-generated voices and human voices really depends on the context and your specific project goals. AI voice over technology has come a long way and can provide natural-sounding voices, but some may still prefer the authenticity and emotion that human voices bring. It's a balance between efficiency and genuine character portrayal. So, the answer might vary from project to project. It's all about what serves your game or content best.

 

[stormluisss]

It’s certainly cool that technology has reached the point of simulating the human voice, but to be honest, I don’t really understand where it can be used usefully other than the voice acting of a translator or Google maps and answering machine. Although, judging by the way technology is developing today, AI will soon be dubbing films. But I’m a man of the old school, and I don’t believe that he can surpass ordinary human talent.

 

[Colif]

AI couldn't really make movies worse than what we had already so I don't see a problem. I don't really play games that have talking in them so I don't really care one way or other.

AI needs to become more intelligent before I worry about it. Right now its mostly just replacing search engines... Google wants to replace search with it, but they stealing content from other sites and presenting it as theirs... bad google.

 

[Frag Maniac]

I feel voice and dialog are definitely two areas where AI should NOT be used. I just don't think a computer generated version is ever going to have the emotional depth and realism. The worst part of it is, if it catches on, it will put good writers and voice actors out of work.

 

[stormluisss]

AI couldn't really make movies worse than what we had already so I don't see a problem. I don't really play games that have talking in them so I don't really care one way or other.

AI needs to become more intelligent before I worry about it. Right now its mostly just replacing search engines... Google wants to replace search with it, but they stealing content from other sites and presenting it as theirs... bad google.

Yes, Google has become quite cunning now. And this voice search already prevents you from simply writing a request. It’s especially awkward when the voice-over of your request comes on somewhere at work in an office full of people.

 

[Colif]

Do you mean https://chrome.google.com/webstore/detail/voice-search/hhfkcobomkalfdlmkongnhnhahkmnaad

Lucky for me my PC is blind and deaf. No camera or Microphone. Damn... can't use voice search... oh well. Glad its not enforced on my phone yet.

 

[stormluisss]

Do you mean https://chrome.google.com/webstore/detail/voice-search/hhfkcobomkalfdlmkongnhnhahkmnaad

Lucky for me my PC is blind and deaf. No camera or Microphone. Damn... can't use voice search... oh well. Glad its not enforced on my phone yet.

Yep, something like that, especially on my phone I have a built-in Google application like Siri. It wedged itself into every conversation I had until I completely lost it. At first I thought it was a cool thing, but considering how much inconvenience it brings, the creators clearly need to improve it

 

[Colif]

I never set them up on my phone. Its there, I just don't use it unless I accidentally press wrong keys. There are advantages to older phones, the new features just don't get ported back. One day I will have to get a new phone but I only use mine for sms and listening to music.

Now I see bing chat appearing in places I don't want it... 1st on PC desktop, now on the keyboard software on my phone. No... just stop. If I want AI help, I will ask. Its pretty useless anyway, they mostly there to sell you things.

 

[neogunhero]

The Finals dev has a very strange response to outrage over its use of AI voices

After drawing criticism online over The Finals using AI voice performances, developer Embark Studios has defended the decision.

www.vg247.com

The Finals uses AI generated voice for their in-game shouts/callouts. They defend it by saying it helps them work faster and provides more contextual voices, which I think they could easily get from voice actors. The AI voice almost seems like it is supposed to be temporary but it’s what they actually wanted.

In this sort of instance, I think it’s acceptable but the voices do sound bad. They are so robotic and static. However, this is a multiplayer FPS game where you are doing these callouts to bring attention to your teammates about different contextual things. If the voices just sounded better and more natural, I wouldn’t mind it. Plus it is based off real voice actors and employee voices, all of which I assume have given consent to this. It seems like it was ethically done at the very least. I just don’t want to see this in story/character heavy games where it is trying to create a real emotional and human connection.

 

[IronDeniskach]

The Finals dev has a very strange response to outrage over its use of AI voices

After drawing criticism online over The Finals using AI voice performances, developer Embark Studios has defended the decision.

www.vg247.com

The Finals uses AI generated voice for their in-game shouts/callouts. They defend it by saying it helps them work faster and provides more contextual voices, which I think they could easily get from voice actors. The AI voice almost seems like it is supposed to be temporary but it’s what they actually wanted.

In this sort of instance, I think it’s acceptable but the voices do sound bad. They are so robotic and static. However, this is a multiplayer FPS game where you are doing these callouts to bring attention to your teammates about different contextual things. If the voices just sounded better and more natural, I wouldn’t mind it. Plus it is based off real voice actors and employee voices, all of which I assume have given consent to this. It seems like it was ethically done at the very least. I just don’t want to see this in story/character heavy games where it is trying to create a real emotional and human connection.

All this said, it's definitely cheaper than paying people to do each voice. And it’s easier in principle when the AI gives you the recording you need ready-made, instead of bothering with recording and editing. But it definitely won’t give the same emotions as a real voice. It’s a pity that they save on our emotions.