Google’s DeepMind Creates Another Real-World Breakthrough In Synthetic Speech Generation

[Lucian Armasu]

Google used its DeepMind AI technology to drastically improve how natural its text-to-speech engine can sound. The project is still only in the research phase for now due to how computationally expensive it is, but it showed great promise.

Google’s DeepMind Creates Another Real-World Breakthrough In Synthetic Speech Generation : Read more

 

[kenjitamura]

Yay. We're now one step closer to making celebrities say whatever the hell we want them to say and floating the clips around the internet for people to think they're real sound bytes.

 

[Pnutt79]

Don't we do that already?

 

[McWhiskey]

So we fully replace actors visually with cgi. Step 2, replacing them audibly with "cga" is right around the corner.

Then actors will be obsolete.

But if actors are obsolete, who will we collectively worship?

 

[bloodroses]

Oh great.... now the spam phone calls we receive from telemarketing machines are going to get more realistic sounding; or tech support calls for that matter.

I guess that means that voice recognition security is now useless as well.... lol

 

[lorfa]

Should have included the link: https://deepmind.com/blog/wavenet-generative-model-raw-audio/

Gives the examples and explains better what it is they're doing.

 

[Flying-Q]

So we fully replace actors visually with cgi. Step 2, replacing them audibly with "cga" is right around the corner.

Then actors will be obsolete.

But if actors are obsolete, who will we collectively worship?

Do bear in mind that there will still need to be directors and animators creating the content as Deep Mind has not yet got to the stage of cohesive dramatic inspiration.

 

[George Phillips]

We are getting closer and closer to creating the robots in the movie A.I. It will definitely come true!

 

[fixxxer113]

It's all good until one day, you hear a perfectly synthesized human voice saying

"I'm sorry Dave, I'm afraid I can't let you do that..."

 

[ccampy]

Voice recognition was pretty bad as security allready

 

[BrownRecluse27]

Wavenet? why not just call it Skynet?

 

[bit_user]

WaveNet is a fully convolutional neural network that can modify the raw waveform of the audio signal one sample at a time.

Saying it "modified" the raw waveform confused me. I think you meant to say "synthesized".

But if actors are obsolete, who will we collectively worship?

I don't know, but I'm really liking this Sharon Apple I've been hearing, lately.

Should have included the link: Was]https://deepmind.com/blog/wavenet-generative-model-raw-audio/

Was that not in the Source: link, under the headline? It's there, now.

I've never seen Lucian fail to source his articles. I also doubt I've seen him reply to any comments (does he even read them?).

Wavenet? why not just call it Skynet?

Well... because it's not?