When I posted about deepfakes a few years ago people wondered when voice will happen well..

PEB

Sunflower in support of Ukraine
Platinum Member
Joined
Jan 20, 2004
Messages
33,501
Reaction score
25,462
Tons of new software for mimicking human voice is coming around with lyrebird and I believe a direct ripoff DeepVoice from China. The are getting impressively close to text to speech voice reporduction of the original source material. It can take 100's of small samples to get it close to the original material but they are getting better. You could with enough samples make a Model or Celebrity talk dirty to you or have President Trump declare war on Russia. Seriously the stuff is getting really good.

https://lyrebird.ai/signup

 
Read in NYT the big tech being rolled out at all the conferences this year is mostly voice activated/related.


Still too paranoid to get Alexa.
 
I don’t see how this is a good thing to unleash on the world. Will make it too easy to impersonate people. Just more technology to help scam artists.
 
Read in NYT the big tech being rolled out at all the conferences this year is mostly voice activated/related.


Still too paranoid to get Alexa.

So your telling instead of talking to Alexa I could give orders to Britney Spears or Taylor Swift in the future. :)
 
I don’t see how this is a good thing to unleash on the world. Will make it too easy to impersonate people. Just more technology to help scam artists.


This is why I was posting this because more people are aware of this stuff the better.
 
Lemme know when they can mimic the voice of my cousins wife so I can stop calling to ask for recipes or awkward life advice

BewitchedSpiritedItalianbrownbear-small.gif
 
Combine this stuff with this stuff you see where it could go and will only get better.

 
I just tried it, it worked....NOT

How many samples did you take? If you just tired it unlikely you did anywhere near the number of samples needed to make it work properly.
 
Yup, this is yet another AI game changer.

Hopefully AI can come up with a way to determine what's fake and what's not... or we potentially face some very, very troubling situations.
 
How many samples did you take? If you just tired it unlikely you did anywhere near the number of samples needed to make it work properly.
35, how many till I dont sound like a drunk asshat?
 
35, how many till I dont sound like a drunk asshat?

Likely not possible then because you seem to sound like a drunk asshat anyways without AI's help.
 
Back
Top