3 ms·
I'd love to read that if you ever put it together! How many hours of audio did it need to train one of the voices?
by lpellis 6y ago
I'd love to read that if you ever put it together!
How many hours of audio did it need to train one of the voices?
- echelon 6y agoIt varied a lot. Some speakers (that sound great) only had about 45 minutes of audio. Other speakers had up to five hours. The key to training is that all of the models were transfer learned from the Linda Johnson speech dataset (LJS).