Synthesizer for Training

1) loads audio, speaker_id, text pairs 2) normalizes text and converts them to sequences of integers 3) computes spectrograms

Dialted and Depth-Separable Convolution

Maintain similar input lengths in a batch. Length groups are specified by boundaries. Ex) boundaries = [b1, b2, b3] -> any batch is inclu

Zero-pads model inputs and targets

Zero-pads model inputs and targets

1) loads audio, text pairs 2) normalizes text and converts them to sequences of integers 3) computes spectrograms from audio