Deep-learning text-to-speech tool for generating voices of various characters

Re:Deep-learning text-to-speech tool for generating voices o... by trowngon on Wednesday January 06, 2021 @04:09AM (#25655967)

Are there open source projects like this?

[ Parent | on HN ]

Re:Deep-learning text-to-speech tool for generating voices o... by CookieAnon on Wednesday January 06, 2021 @04:30AM (#25656069)

I have CookieTTS where I reseach lots of experimental stuff. (You can see my credits on the 'Thanks' section of 15.ai)

I can get about 90% of the quality of 15.ai currently. I think I could surpass 15.ai but not without some help.

[ Parent | on HN ]

Re:Deep-learning text-to-speech tool for generating voices o... by nmstoker on Wednesday January 06, 2021 @06:34PM (#25664314)

There's Mozilla TTS https://github.com/mozilla/TTS

Here's a sample from a TTS model + vocoder I released for it. I've no wish to deter the motivated, but it'd take a bit of figuring out how to set things up and you'd need to read the docs and code to get oriented :)

https://m.soundcloud.com/user-726556259/sherlock-wavegrad-sa...

Links to the models are here: https://discourse.mozilla.org/t/creating-a-github-page-for-h...

Is originally trained on two novels read by the same narrator on LibriVox (ie in public domain)

[ Parent | on HN ]

Re:Deep-learning text-to-speech tool for generating voices o... by nmfisher on Thursday January 07, 2021 @10:22AM (#25672052)

This is actually quite impressive too, significantly better than the last time I looked into Mozilla TTS. Roughly how much audio does "two novels" equate to?

[ Parent | on HN ]

Re:Deep-learning text-to-speech tool for generating voices o... by nmstoker on Thursday January 07, 2021 @06:37PM (#25679220)

Here's another sample with the same model+vocoder, this time reading from a Wikipedia article: https://m.soundcloud.com/user-726556259/q-learning-wavegrad-...

[ Parent | on HN ]

Re:Deep-learning text-to-speech tool for generating voices o... by nmstoker on Thursday January 07, 2021 @06:26PM (#25679107)

It's about 32 hours of audio.

As some of the audio is read in different accents to the main accent used, ideally the different accent audio would have been removed. Doing so would be expected to help with voice quality, reducing the overall amount used and, as a bonus, cutting training time too.

[ Parent | on HN ]

Re:Deep-learning text-to-speech tool for generating voices o... by terrycody on Wednesday January 06, 2021 @08:33PM (#25666057)

Is there a simple interface like the example in this thread to use the tool for a non developer regard Mozila TTS yet? I can't find one...

[ Parent | on HN ]

Re:Deep-learning text-to-speech tool for generating voices o... by nmstoker on Wednesday January 06, 2021 @10:22PM (#25666889)

There's the demo server which has a simple web UI where you can input text to be spoken, but in regards to setting it up locally it's not that suited for a non developer

https://github.com/mozilla/TTS/tree/master/TTS/server

https://github.com/mozilla/TTS/wiki/Build-instructions-for-s...

There's also a version in docker: https://github.com/synesthesiam/docker-mozillatts

And various Colabs too, which are fairly easy to get going with: https://github.com/mozilla/TTS/wiki/TTS-Notebooks-and-Tutori...

[ Parent | on HN ]