New AI able to clone any voice with just seconds of input data

[RejZoR]

I love the use of "Ai" again. Like for 99% of things marketed. There is no "Ai". Our voice literally consists of tone and pitch. What they do is literally analyze these parameters and thy "clone" your voice using "Ai". God, it sounds so funny as they desperately shove "Ai" buzzword into everything these days... There's no Ai, they just shoved voice recording through what we'd call an audio analyzer back when buzzwording was still normal...

[laminutederire]

6 hours ago, BuckGup said:

All I saw was a link to code that was made independently of the researchers that was similarly based in how it worked

More than half of that kind of papers aren't reproducible anyway because people don't share their data sets. That's usually how it goes with google code anyway.

I would be very critical of those results solely for that. It just displays the usefulness of techniques developed in other fields, and if those other fields are of any use, the results are probably limited. Just saying

[Doobeedoo]

Spoofing aside. Everyone a vocaloid, easy. 

[vanished]

10 hours ago, RejZoR said:

I love the use of "Ai" again. Like for 99% of things marketed. There is no "Ai". Our voice literally consists of tone and pitch. What they do is literally analyze these parameters and thy "clone" your voice using "Ai". God, it sounds so funny as they desperately shove "Ai" buzzword into everything these days... There's no Ai, they just shoved voice recording through what we'd call an audio analyzer back when buzzwording was still normal...

It uses machine learning and neural networks to create the audio analyser that can then perform the "clone".  I know the term is often misused these days, but a) that's nothing new - we've been calling hard coded logic in video game npcs ai for decades now, and b) I think by any common definition this thing is true AI.

[Unixsystem]

Ignoring the much more serious potential negative consequences of this tech, I see huge potential in this sort of AI for game development. While bigger studios can eat the cost of hiring dozens of voice actors for your average AAA game, this sort of generated voice could be huge for smaller devs making RPGs or other dialogue-heavy games. Even more so when you factor in the longer term cost of maintaining voice actors for certain roles.

 

If a character-based shooter wants to add a new map, they may need to track down every voice actor for every character to record more callous or map-specific lines (or omit that sort of dialogue altogether). Same deal for a sequel, except on an even bigger level. With this they would just need a folder of presets and be good to go. 

 

It may not be great news for the voice acting community long term, but chances are that it'll be a while before the tech is really up to snuff and indistinguishable across a wide range of emotions and situations, but still really promising as a way to streamline voice recording. 

[vanished]

39 minutes ago, Waffles13 said:

Ignoring the much more serious potential negative consequences of this tech, I see huge potential in this sort of AI for game development. While bigger studios can eat the cost of hiring dozens of voice actors for your average AAA game, this sort of generated voice could be huge for smaller devs making RPGs or other dialogue-heavy games. Even more so when you factor in the longer term cost of maintaining voice actors for certain roles.

 

If a character-based shooter wants to add a new map, they may need to track down every voice actor for every character to record more callous or map-specific lines (or omit that sort of dialogue altogether). Same deal for a sequel, except on an even bigger level. With this they would just need a folder of presets and be good to go. 

 

It may not be great news for the voice acting community long term, but chances are that it'll be a while before the tech is really up to snuff and indistinguishable across a wide range of emotions and situations, but still really promising as a way to streamline voice recording. 

It will be interesting to see how they respond to the tech if and when it really starts getting used this way.  I suspect contracts will stipulate that they must still be paid for using the likeness of their voice in such a manner, otherwise as you say, it could really destroy the industry.

[Unixsystem]

1 hour ago, Ryan_Vickers said:

It will be interesting to see how they respond to the tech if and when it really starts getting used this way.  I suspect contracts will stipulate that they must still be paid for using the likeness of their voice in such a manner, otherwise as you say, it could really destroy the industry.

If I was a voice actor right now, I would probably be busting my butt looking for regular acting roles. 

 

I don't have anything against voice actors in general, but if the tech gets good enough to do their job with basically no overhead and way faster turnaround, then I can't say I'll be too upset if their industry largely goes away. The same way I'm not too upset at the lack of a switchboard operator industry today, or the burger flipping industry will be in 5-10 years. 

[vanished]

Just now, Waffles13 said:

If I was a voice actor right now, I would probably be busting my butt looking for regular acting roles. 

 

I don't have anything against voice actors in general, but if the tech gets good enough to do their job with basically no overhead and way faster turnaround, then I can't say I'll be too upset if their industry largely goes away. The same way I'm not too upset at the lack of a switchboard operator industry today, or the burger flipping industry will be in 5-10 years. 

Yeah I'm not saying we need to ignore technology in order to preserve it at all costs or anything like that.  If we did that we'd still be communicating with cave paintings lol

I just expect there to be push back, and tbh I don't think it would even be unreasonable for them to do so.

[Unixsystem]

6 minutes ago, Ryan_Vickers said:

I just expect there to be push back, and tbh I don't think it would even be unreasonable for them to do so.

Fully agreed, but regardless of what they try to do to stay relevant, I would hope that most voice actors are smart enough to see the writing on the wall. Even a year or two ago voice fonts were insanely expense and time consuming to produce, so at the rate we are progressing I can't imagine that it'll be more than a decade before we start seeing widespread adoption of artifical voice acting in games/animation. 

 

Spoiler

Add in further improvements to machine translation and suddenly anime becomes way more accessible as well.