2017-04-26 03:09:59

hey I just heard about this new lyrebird tts. Its still in development. But apparently you can take a 60 second recording of a person and using just 1 minute of their voice, make a tts with it. here is a link to an extensive article on the topic. if you also google search lyrebird tts you can learn more. the link is

https://techcrunch.com/2017/04/25/lyreb … -news-era/

let me know what you all think of this. I for one think its excellent news for us.

2017-04-26 17:13:57

Whoa! This stuff is cool. NOw once it reaches consumer's hands, just watch what'll happen with it. Part of me wants to know what would happen if this was combined with Wavenet. Although I'm pretty sure this thing has neural networks in place already, the naturalness of wavenet will take quite a while to beat.

Make more of less, that way you won't make less of more!
If you like what you're reading, please give a thumbs-up.

2017-04-26 18:16:31

this is really cool. it's definatly going to be less cumbersome than model talker and mnlp, both of which required a large number of recordings and were bearly accessible (modeltalker especially). I'm probably going to put a 60 second sample of gnuspeech onto this beauty, and maybe create tts simulating the older pre-dectalk synthesyzers (davo, patern playback...)

be a hero and stop Coppa now!
https://docs.google.com/document/d/1Dkm … DkWZ8/edit
-id software, 1995

2017-04-26 19:49:54

Reminds me of this. Fun stuff.

To see a world in a grain of sand, and a heaven in a wild flower.
Hold infinity in the palm of your hand, and eternity in an hour.
William Blake - Auguries of Innocence, line 1 to 4

2017-04-27 03:17:54

This looks awesome! I hope its accessible.

Kingdom of Loathing name JB77

2017-04-27 06:15:49

Whoa! And I thought Inoetics's few hours of samples with your own script was as far as it could get! lol That's awesome! I could see tons of use coming out of making a synth is 60 seconds. You could, in addition to your voice, emulate dozens of other synths. Sure, it'd probably be online only and it wouldn't sound exactly the same if synth emulation is your thing, but it's seriously got potential.

2017-04-27 18:33:09

speaking of wavenet, what is the future of it?

be a hero and stop Coppa now!
https://docs.google.com/document/d/1Dkm … DkWZ8/edit
-id software, 1995

2017-04-27 23:22:17

Looks like it's making its way into google home, they have a prototype that can generate synthesis samples as well as do some pretty weird things to music thrown at it.

2017-04-28 07:45:38

I hope wavenet will be the future of speech synthesis in the 2030s. Well, I think so when the Microsoft voices were upgraded to the WN quality.

73 Wj3u

2017-04-28 21:38:13

It would be interesting what would happen if you tossed a music track at this thing. I wish we could test this API, but I'm guessing we'll have to wait until early next year or late 2017. All the same, it sounds pretty awesome.

Discord: dangero#0750
Steam: dangero2000
TWITCH
YOUTUBE and YOUTUBE DISCORD SERVER

2017-04-29 05:51:57

microsoft voices upgraded to the wn quality? what is wn? how do i get such voices?

2017-04-30 15:42:30

It's not true. He got confused. Microsoft voices are not, nor will they ever be, upgraded to Wavenet. Unless they released Wavenet as an sdk for devs, which, hate to reign in on the  parade, but I highly doubt this would happen. Even if they really wanted to, they couldn't. They would have to release the whole damn neural network, which would not be possible. They have ridiculously unbelievable amounts of compute power allocated to this. Not sure on exact stats, but just know this, their compute power would make 8gb of ram look like 256mb compared to this.

2017-04-30 17:21:28

Does lyrebird have their own website yet?

Discord: dangero#0750
Steam: dangero2000
TWITCH
YOUTUBE and YOUTUBE DISCORD SERVER

2017-04-30 18:58:32

Yes they do here is their website.

https://lyrebird.ai/

Kingdom of Loathing name JB77

2017-05-01 16:00:43

Hi people!
I have heard about It too and I must say, this is very cool. I would like to request those people to apply another languages to lirebirds too.
But, I don't know, that this is concatenative, or formant process.

Ja volim samo kafu sa Rakijom.

2017-05-01 19:06:07

I can apply spanish language unless they make the voice of the president of Panama...
As I was saying, WN is short for wavenet, unless I comfuced you.

73 Wj3u

2017-05-03 00:53:05

Yes, and Wavenet is Google's endeavor, no one elses.

2017-05-04 02:19:51

do you want to play with wavenet for free on your computers? well here you go. here is the website for google's open source wavenet.

https://github.com/ibab/tensorflow-wavenet

2017-05-04 04:57:19

Well, unless I say this it's illegal.

73 Wj3u

2017-05-04 23:55:12

It's not illegal, but I have a feeling that may be just the charts and not actually the speech synthesis. Unless it's a gui frontend to the neural network.