Support NeoGAF

[Mikey Jr.]

So I'm playing the JRPG Trails of Cold Steel 4 right now, and while I love the game, I realize that these games can't have huge budgets. They don't sell a lot, so putting money into lots of VO probably isn't a wise decision.

I would say only about 20% of the time, the main characters speak. Usually during important events. The rest is just text. Side characters

And while I love the characters a lot, no VO kinda sucks.

And then I stumbled on AI Voice. Its mainly just text to speech, but it tries to mimic real human voice.

Now I'm not gonna sit here and tell you that what they have right now is perfect. You can tell its a robot, but at its current point, its pretty damn good.

And it got me thinking. While big budget games can go with real actors, these smaller games can utilize AI. And so now every game in the series can have the same character voices between games. No "ohhh the actor wasn't available" or "we couldn't pay them enough" or "we only had enough budget for these lines".

The entire game could be completely voiced. Even NPC's on the side of the road could have their own voices.

Is this feasible right now? Nahhh. The voices aren't perfect yet. But in 10 years?

What do you guys think?

 

[jshackles]

That's definitely the way things are moving. For example - this project is being used to do some really impressive stuff with The Elder Scrolls games, all using AI synthesized voices. It's only a matter of time before something like this is made available as a tool for lower budget games.

And I'm with you, I think it'd be great.

 

[lestar]

It's a matter of time an easy to use and cheap/free tool is available for everyone, currently there're github projects and paid professional tools that use AI synthesized voices

 

[reksveks]

Adobe's VoCo voice project: Now you really can put words in someone else's mouth

It may be hard to trust the authenticity of any recorded speech in the not too distant future.

www.zdnet.com

I see something like this being used in the future. It primarily going to be used in marketing and movie/film entertainment (aka your favourite actor speaking in a different language) but it will trickle down to games.

 

[Belmonte]

Alien Fires: 2199 AD.

Made in 1986, 2 floopy discs for Amiga.

 

[Magic Carpet]

Thats all folks. We can fire all the voice actors now.

 

[CamHostage]

Adobe's VoCo voice project: Now you really can put words in someone else's mouth

It may be hard to trust the authenticity of any recorded speech in the not too distant future.

www.zdnet.com

I see something like this being used in the future. It primarily going to be used in marketing and movie/film entertainment (aka your favourite actor speaking in a different language) but it will trickle down to games.

Right, the technology is out there, and it's getting smoother and more realistic all the time. A type of "Voice Deepfake" is coming.



I don't know if voice actors will be lining up to provide their voices for a scan and let the producers give only 10 minutes of work to crank out a full production... but as shown in even Adobe's "fun" demo of VoCo, voice actors may not have much choice.

Unreal Engine plugins exist for turning text into speech in the voice of a dedicated character. They can read just flat text and that comes out as expected, but also now, that text can be fine-tuned with sliders and markers on sections of the lines to adjust emphasis and emotion. I've not seen a game go this route seriously yet (there have been text-to-speech options in games a few times I believe, but so far the artificiality is just too imposing; these Japanese RPGs particularly have a certain style to the characters that I have a feeling the creators would prefer you to experience with your imagination while reading,) but the technology and the level of control is getting better.

 

[nkarafo]

Yeah, no.

I think plenty of voice actors do a bad enough job already. You are now asking for some robot to do an even worse one. That's a sure way to ruin one game's atmosphere.

Unless the NPCs are robots. I can see that working in a Portal game. Then again, Portal already has full voice acting so...

 

[reksveks]

Right, the technology is out there, and it's getting smoother and more realistic all the time. A type of "Voice Deepfake" is coming.



I don't know if voice actors will be lining up to provide their voices for a scan and let the producers give only 10 minutes of work to crank out a full production... but as shown in even Adobe's "fun" demo of VoCo, voice actors may not have much choice.

Think voice actors will have to sign 'better' contracts with more detailed terms but yeah there will be some resistance there. I would like more fully voiced games so okay with whatever makes it more feasible, hopefully devs give me a choice.

 

[CamHostage]

Yeah, no.

I think plenty of voice actors do a bad enough job already. You are now asking for some robot to do an even worse one. That's a sure way to ruin one game's atmosphere.

Unless the NPCs are robots. I can see that working in a Portal game. Then again, Portal already has full voice acting so...

I don't think we're talking about replacing ALL voice acting (though when the prices come down to almost nothing, it'll be tough for voice actors to picket against "robot scabs".) There's still so, so, so much that's brought to a character thanks to the performance of a voice actor.

For townsfolk in an RPG, though, or all those long text exchanges with stillframe characters in Japanese games (though they'd have to come up with a "kawaii meter" to tune up the voices...), for indie games and titles that cannot afford VO (and maybe in some cases, the Voice AI will replace the language dub for smaller territories, using the same kind of AI pattern recognition/replication that they're using in images and video to transform the English VO into like Portuguese or German?), the text-to-speech is an option where professional VO may not be possible.

You'd still pay Patrick Stewart to belt out five majestic lines and then die, because a computer can only pretend to be Patrick Stewart so well, but for the Blacksmith of Nofuqsberg, you can have an AI say, "Welcome, what are ye' buyin'?" and all that patter text. And it'll go from there...

 

[cormack12]

It doesn't even need to be that complex, this is what WD:Legion did

Watch Dogs Legion has "20 different versions of the script" so you might never get the same cinematic twice

Get ready to sign your life away to endless playthroughs of Watch Dogs Legion

www.gamesradar.com

We’re using technology to alter all the voices so that even if you happen to recruit Joe and John and they have the same actor and the same personality when they talk to each other, you won’t know it’s the same actor because we modulate the voices.

 

[nkarafo]

I don't think we're talking about replacing ALL voice acting (though when the prices come down to almost nothing, it'll be tough for voice actors to picket against "robot scabs".) There's still so, so, so much that's brought to a character thanks to the performance of a voice actor.

For townsfolk in an RPG, though, or all those long text exchanges with stillframe characters in Japanese games (though they'd have to come up with a "kawaii meter" to tune up the voices...), for indie games and titles that cannot afford VO (and maybe in some cases, the Voice AI will replace the language dub for smaller territories, using the same kind of AI pattern recognition/replication that they're using in images and video to transform the English VO into like Portuguese or German?), the text-to-speech is an option where professional VO may not be possible.

You'd still pay Patrick Stewart to belt out five majestic lines and then die, because a computer can only pretend to be Patrick Stewart so well, but for the Blacksmith of Nofuqsberg, you can have an AI say, "Welcome, what are ye' buyin'?" and all that patter text. And it'll go from there...

I honestly prefer the 3rd option. Text.

But i guess it's fine if it's optional.

 

[NahaNago]

This is something I've been thinking about as well. The number of npcs in games are growing larger and larger so hiring enough voice actors for possibly thousands of characters is just not going to work. The key here is when you write the dialogue for the characters you also code the way the person says the words at the same time. Keep voice actors for the main storyline characters and some side quests.

also randomly generate furnishing apartments and homes for cities and a.i. citizen lifestyle for cities to create the atmosphere of folks living.

 

[DevilHunterX0]

You mean like Operation Stealth on Amiga?

 

[Soodanim]

That's definitely the way things are moving. For example - this project is being used to do some really impressive stuff with The Elder Scrolls games, all using AI synthesized voices. It's only a matter of time before something like this is made available as a tool for lower budget games.

And I'm with you, I think it'd be great.

First post nailed it. When this started happening I realised it won't be long until this tech is developed and sold, and it's going to be useful for smaller companies. I don't think there's any danger of bigger games giving up on actual human voices, that would require truly perfect synthesis and that's probably a long way off.

A perfect early use case would be an indie studio to make a game about robots, to use the infancy of the tech to their advantage.

 

[The Cockatrice]

Nah Im good. Just hire a real person. I'd rather have some emotional impact than a robotic voice. I've seen how it was done in WDL and that was with huge budget, indies will never reach that high and it was absolutely shit in WDL.

 

[martino]

Nice solutions in here.
Great for indies
i love how the tech and tools progress is slowly giving more freedom and capabilities

 

[RedVIper]

Yeah no, I'd rather read text and make my own interpretation of what's being said than have shit voice acting, it's already shit enough, I've yet to see a game that matches VtM:Bloodlines of all things and that games is old as fuck.

It could be used for random NPC's, like when you press E and they say a random line, that could be okay, but main characters would still require actual voice acting, and small projects can't afford that.

Bad va really pulls me out of games so I hope this doesn't become a thing.

 

[stranno]

I'm fine with gibberish language, tbh, it works fine in most games.