2 ms·
If I understood it correctly, the datasets being used are all in the order of ~20 hours. Could the results be improved with a different dataset? Say, Mozilla's
by rglullis 9y ago
If I understood it correctly, the datasets being used are all in the order of ~20 hours. Could the results be improved with a different dataset? Say, Mozilla's Common Voice[1]? They are already at 350 hours of labeled speech.
[1]: https://voice.mozilla.org https://voice.mozilla.org
- IshKebab 9y agoMaybe, but Google's tacotron2 also used about 24 hours of speech and gives results that are indistinguishable from humans.