Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

StreamVC: Real-Time Low-Latency Voice Conversion

research.google

Post reply on HN
Print view View original

31–40 of 44 posts

Re: StreamVC: Real-Time Low-Latency Voice Conversion

#31

Post by huac »

kaliqt wrote: post #21
refulgentis wrote: post #14

Earlier quoted context omitted.

> But I suppose that would be too controversial for Google. You don't have to suppose anything: it is actually settled law that its bad to just willy-nilly use people's voices if you feel like it, even if its just a sound-alike!

Sound-alike is not actionable, and anyone who claims it infringes should be dragged through the courts. Not your voice, not your IP, not your right.

except to the extent that your voice may be part of your image, which is actionable: https://en.wikipedia.org/wiki/Midler_v._Ford_Motor_Co.

Re: StreamVC: Real-Time Low-Latency Voice Conversion

#32

Post by TylerE »

manishsharan wrote: post #24

Are there any use cases that is driving this ? Is there a huge burning need for technology ? Are kidnappers and con-men a huge under-served market that Google is hoping to serve ? Deep Fake videos not convincing enough to serve the need of fraudsters ? I am totally against regulating AI but shit like this gives fodder to the other side.

Since the quality is pretty listenable, one use case I can see is youtubers who want to do voiceovers on their videos, but not be linked to their real world identity.

Heck, I can even see broadcasting uses. Imagine if every on-air personality had good target files made ahead of time, so then when they catch a cold, production runs their lapel mic feed through this, using the "good" target sample, and remove all the congestion and raspiness.

Re: StreamVC: Real-Time Low-Latency Voice Conversion

#33

Post by pessimizer »

manishsharan wrote: post #24

Are there any use cases that is driving this ? Is there a huge burning need for technology ? Are kidnappers and con-men a huge under-served market that Google is hoping to serve ? Deep Fake videos not convincing enough to serve the need of fraudsters ? I am totally against regulating AI but shit like this gives fodder to the other side.

You're totally against regulating AI, but the idea that AI could aid anonymity makes you want to regulate AI?

> I am totally against regulating AI but shit like this gives fodder to the other side.

You think anonymity is so universally hated that it's actually bad PR for leaving AI completely unregulated? No other problems with AI that you can think of, and also no good reason why someone should be allowed to be anonymous?

Re: StreamVC: Real-Time Low-Latency Voice Conversion

#35

Post by modeless »

refulgentis wrote: post #14
modeless wrote: post #6

Earlier quoted context omitted.

Not a very good demo page. It's difficult to judge real world quality with such unenthusiastic reading, unrealistic sentences, and unfamiliar voices. Typical of speech papers. It would be much better if celebrities were used as target voices, as we all know what they sound like and can therefore judge quality better. But I suppose that would be too controversial for Google. In general I think it is silly that voice c…

> But I suppose that would be too controversial for Google. You don't have to suppose anything: it is actually settled law that its bad to just willy-nilly use people's voices if you feel like it, even if its just a sound-alike!

So it's illegal for standup comics to do impressions? Or maybe there is some nuance here?

Re: StreamVC: Real-Time Low-Latency Voice Conversion

#36

Post by swatcoder »

neilk wrote: post #22

What are the anticipated use cases? I know of one: transgender people often would like to alter the timbre of their voice and spend a lot of time training their voice. At least for online scenarios, this can just do it. But other than that AI voice altering research seems like it benefits mostly scammers? I’m just wondering what they tell themselves they’re doing. I didn’t see this in the paper.

I think it's hard to see the use case right now because the quality remains pretty dreadful.

But the prototypical legitimate use case (which we needn't be excited about), is a voice over artist leasing their timbre instead of their time so that new text can be made to sound like them without their being actively involved. If it were to become mature (which doesn't seem close, from this example), it would be a big step up from existing phone tree voice assemblage and would open the doors for dubbing, animation voiceover, harmonization, and ADR in commercial sound and film.

Gender masking or general anonymization aren't really served by this, as you don't need to adopt a specific target timbre to deliver on those. There are other techniques that work perfectly well for those uses, some that have already been around for ages.

Re: StreamVC: Real-Time Low-Latency Voice Conversion

#37

Post by nine_k »

refulgentis wrote: post #14
modeless wrote: post #6

Earlier quoted context omitted.

Not a very good demo page. It's difficult to judge real world quality with such unenthusiastic reading, unrealistic sentences, and unfamiliar voices. Typical of speech papers. It would be much better if celebrities were used as target voices, as we all know what they sound like and can therefore judge quality better. But I suppose that would be too controversial for Google. In general I think it is silly that voice c…

> But I suppose that would be too controversial for Google. You don't have to suppose anything: it is actually settled law that its bad to just willy-nilly use people's voices if you feel like it, even if its just a sound-alike!

Could two researchers with different voices and manners of speech agree to provide longer samples of their voices? Then they would convert each sample to imitate the other's voice. It would be easy to contrast and compare, without any controversy.

Re: StreamVC: Real-Time Low-Latency Voice Conversion

#38

Post by Dracophoenix »

neilk wrote: post #22

What are the anticipated use cases? I know of one: transgender people often would like to alter the timbre of their voice and spend a lot of time training their voice. At least for online scenarios, this can just do it. But other than that AI voice altering research seems like it benefits mostly scammers? I’m just wondering what they tell themselves they’re doing. I didn’t see this in the paper.

Voiceover/broadcasting. Recording or acquiring any audio that isn't freely licensed background music is among the most expensive and time consuming parts of a prerecorded broadcast. With voice alteration, a director and sound engineer can become their own actors in anything ranging from commercial spots to large-scale and long-running animated shows.

Re: StreamVC: Real-Time Low-Latency Voice Conversion

#39

Post by refulgentis »

modeless wrote: post #35
refulgentis wrote: post #14

Earlier quoted context omitted.

> But I suppose that would be too controversial for Google. You don't have to suppose anything: it is actually settled law that its bad to just willy-nilly use people's voices if you feel like it, even if its just a sound-alike!

So it's illegal for standup comics to do impressions? Or maybe there is some nuance here?

Thanks modeless, I appreciate the comment -- yes, I agree, there is a substantive difference between "us[ing] people's voices" and "[doing an] impression [of someone's voice]". It's good you pointed this out, law can be hard for swes, it can tickle the "unambiguous rule" part of the brain we employ every day.

Re: StreamVC: Real-Time Low-Latency Voice Conversion

#40

Post by neilk »

thinkski wrote: post #25
neilk wrote: post #22

What are the anticipated use cases? I know of one: transgender people often would like to alter the timbre of their voice and spend a lot of time training their voice. At least for online scenarios, this can just do it. But other than that AI voice altering research seems like it benefits mostly scammers? I’m just wondering what they tell themselves they’re doing. I didn’t see this in the paper.

I suspect one is masking that a call center is in a low wage country, e.g. make customer in U.S. believe they’re talking to someone in U.S. while paying a fraction of the U.S. wage.

Right. I thought of that too, but it doesn’t mask accents, at least not yet

I suppose if you could make agents all sound the same they would be interchangeable, and companies always love that. It’s Anjali or Ligaya or Dolores but now they all sound like “Becky”?

Post reply on HN