News for hackers. Stuff that matters.

A comment on An open source implementation of DeepVoice 3: 2000-Speaker Neural Text-to-Speech.

Re:An open source implementation of DeepVoice 3: 2000-Speake... by dspig on Thursday April 26, 2018 @05:20AM (#16930273)
The "chirping" artifacts are quite distracting. I wonder if they can be avoided by randomizing the alignment of the final piecing together of the audio?

Post-processing by convolving with a short noise burst removes them pretty well, as that randomizes phase vs. frequency.

[ Parent | on HN ]
Re:An open source implementation of DeepVoice 3: 2000-Speake... by _bxg1 on Thursday April 26, 2018 @02:16PM (#16934217)
I was surprised that the inflection was so good but the chirping was so bad. I'm no expert, but the latter seems like a much less difficult problem.
[ Parent | on HN ]
Re:An open source implementation of DeepVoice 3: 2000-Speake... by joshumax on Thursday April 26, 2018 @01:23PM (#16933794)
IIRC, I think wavenet deals with residual hissing and chirping through a causal dilated convolution layer.
[ Parent | on HN ]
Re:An open source implementation of DeepVoice 3: 2000-Speake... by skykooler on Thursday April 26, 2018 @12:22PM (#16933295)
Would it be possible to upload the outputs of this convolution? I'd like to hear how it sounds.
[ Parent | on HN ]
Re:An open source implementation of DeepVoice 3: 2000-Speake... by dspig on Thursday April 26, 2018 @01:55PM (#16934044)
Here's the second example convolved with 30ms white noise:

https://drive.google.com/file/d/1WwmCNwWwukhtYXRT4mDHxCUa0sQ...

But it also makes it sound like she's in a closet, so it would be better to fix it at source if there's a way to do that.

[ Parent | on HN ]