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

Lyrebird – An API to copy the voice of anyone

lyrebird.ai

Post reply on HN
Print view View original

51–60 of 311 posts

Re: Lyrebird – An API to copy the voice of anyone

#51

Post by bambax »

ehsankia wrote: post #18
yladiz wrote: post #6

This is pretty cool (although, I have no idea what other technologies exist for this kind of thing), but it's definitely not convincing enough to a human listener. This sounds like it might be convincing enough for some programs like "Hey, Siri" but it's not gonna convince your mom. You can listen to the samples on the page linked here and you can immediately tell that Obama and Trump don't sound quite human.

Well, the question is, do they just need to throw more computational power / training at this algorithm or is that the peak of their implementation? This is something Google has been working a lot on [1] and Baidu also recently posted about their results too [2]. We're definitely pretty close to passing the human detectable level. [1] https://deepmind.com/blog/wavenet-generative-model-raw-audio... [2] http://research…

Maybe. However, most CGI (in big Hollywood productions with almost no budget limit) is still very detectable, so much so that the biggest productions try to do as much as possible with real props and real actors instead of CGI (cf. interviews about the last Fast and Furious movie).

The human mind seems to be better at this than most creators credit it for.

Re: Lyrebird – An API to copy the voice of anyone

#52

Post by simcop2387 »

carlob wrote: post #29

I wonder how dependent this is on language: can we make Trump speak Chinese using a one minute audio track of him speaking English?

I'd imagine you might get close, but it'd probably work best if you can get the person to use all the phonemes that you want to reproduce. That said depending on how good it is, it might slur other phonemes together to approximate it, which would probably work to give it the accent that the speaker would likely have.

Re: Lyrebird – An API to copy the voice of anyone

#53

Post by simcop2387 »

celticninja wrote: post #2

Is this enough to beat voice recognition software? If you thought fake news was bad before wait until these 'secret' recordings start getting released and reported on.

That was one of my first thoughts too [1]. I doubt it currently will do so as it does sound fairly robotic. It does sound closer however than other demos I've heard so i think we're probably on the way.

[1] https://youtu.be/-zVgWpVXb64?t=41

Re: Lyrebird – An API to copy the voice of anyone

#54

Post by techdragon »

throwaway29292 wrote: post #36
brango wrote: post #21

Earlier quoted context omitted.

Wow that's amazing. Combine the two and soon Hollywood stars will be redundant. Faceless session actors could just manipulate models of real people who've signed release forms with the vocal performed by similarly faceless vocal artists, or maybe even AI generated voices. Actors would lose their uniqueness and so end up being paid a pittance instead of being able to command the vast sums they can today. Another set o…

I wouldn't be comfortable watching a movie scene if I knew I was looking at computer-generated faces and voices.

Are you comfortable with Auto-Tune in music, not the t-pain / etc exaggerated style... the nearly universal application of Auto-Tune to recording and live performance to ensure a "consistent product", and "save on expensive studio time"?

Because the market appears to have spoken on that one and it said "meh, I don't care" with an solid shrug of indifference.

By the same logic, one can see artificially produced vocal performance combined with artificial overlaying of photorealistic 3d reproduction as a way to cost effectively maximise the performer and crew expenses, and ensure the consistency of a performance. The results may even be better than what they could have done with the real performer in the case of some attractive actors who are not very good at the acting part of being an actor/modern celebrity.

That said, I'll definitely miss the days people had to actually be able to act, but then again I also miss the days people used to actually be able to play an instrument well and or sing well if they wanted to be a famous musician.

Re: Lyrebird – An API to copy the voice of anyone

#55

Post by TeMPOraL »

sudhirj wrote: post #39
amarant wrote: post #19

It's there any copyright protections for a person's voice? If not, David Attenborough and Morgan Freeman will be lead voice actors in my next game project

I don't think you can claim that your game was voiced by either of them, but I don't see how using this would be any more infringing than using a tuned synthesizer.

He probably won't be able to claim the game is "voiced by", but maybe he can get away with saying it features "voices of"?

Re: Lyrebird – An API to copy the voice of anyone

#56

Post by JustFinishedBSG »

backpropaganda wrote: post #34
JustFinishedBSG wrote: post #32

Cooler : http://www.dtic.upf.edu/~mblaauw/IS2017_NPSS/ https://arxiv.org/abs/1704.03809

Canned voices? Robotic intonation? No thanks.

Did you even listen to OP voices and compare them with for example the article Spanish generated voice?

Re: Lyrebird – An API to copy the voice of anyone

#57

Post by zdkl »

sudhirj wrote: post #39
amarant wrote: post #19

It's there any copyright protections for a person's voice? If not, David Attenborough and Morgan Freeman will be lead voice actors in my next game project

I don't think you can claim that your game was voiced by either of them, but I don't see how using this would be any more infringing than using a tuned synthesizer.

Their lawyers > your lawyers though

Re: Lyrebird – An API to copy the voice of anyone

#58

Post by TeMPOraL »

simcop2387 wrote: post #53
celticninja wrote: post #2

Is this enough to beat voice recognition software? If you thought fake news was bad before wait until these 'secret' recordings start getting released and reported on.

That was one of my first thoughts too [1]. I doubt it currently will do so as it does sound fairly robotic. It does sound closer however than other demos I've heard so i think we're probably on the way. [1] https://youtu.be/-zVgWpVXb64?t=41

Throw some hard compression on it to get additional distortion, and I feel you could get away with claiming it was a recording done by a low-power bug.

Re: Lyrebird – An API to copy the voice of anyone

#59

Post by qeternity »

pluma wrote: post #49
qeternity wrote: post #45

While all of these vec2speech type models are impressive, I get the feeling that most of the comments didn't listen to any of the samples. It's still distinctly robotic sounding, probably has quite a bit of garbage output that needs to be filtered manually (as many of these nets often have) and is a far cry from fooling a human.

Sure, it's not clean and crisp but if you ignore the distortions and the somewhat arbitrary intonation it's pretty close to the real thing. And I would argue that it's already good enough to fool grandma for a phone scam (but then again scammers currently get away with practically no resemblance at all).

Scammers have much more success phoning Granny and claiming that they are Windows Support or "The Internet" and simply asking for what they want. There are far easier ways to con than this, and for very sensitive scams where you need spear phishing of this order, then this will fall flat on its face. "Good enough" doesn't quite cut it when it comes to this.

And fwiw, I think the intonations are actually impressively learned and not random. Trump's odd yet distinct intonations are capture quite well in their samples.

Re: Lyrebird – An API to copy the voice of anyone

#60

Post by arghwhat »

qeternity wrote: post #45

While all of these vec2speech type models are impressive, I get the feeling that most of the comments didn't listen to any of the samples. It's still distinctly robotic sounding, probably has quite a bit of garbage output that needs to be filtered manually (as many of these nets often have) and is a far cry from fooling a human.

Most of them are extremely mechanic, to the point where it's almost impossible to understand, but others are actually quite convincing.

I think it primarily needs to learn to respect punctuation, and to translate them to a breathing pause that matches the target voice ("President having speech"-style long pauses vs. "Politician having their ass handed to them by journalist on TV"-no-air-needed pauses).

Post reply on HN