search

descriptinc / melgan-neurips

GAN-based Mel-Spectrogram Inversion Network for Text-to-Speech Synthesis
MIT License
964 stars 214 forks source link

Sample training dataset #2

Closed ghost closed 4 years ago

ghost commented 4 years ago

@ritheshkumar95 A sample dataset to test if training works.

binarythinktank commented 4 years ago

i would be interested in this too, especially to understand how much training wavs/minutes of speaking are actually needed for an accurate result

ritheshkumar95 commented 4 years ago

This is a popular example dataset (also used in this paper) - https://keithito.com/LJ-Speech-Dataset.

You could download this and extract it, and follow the instructions mentioned

binarythinktank commented 4 years ago

perfect, thank you @ritheshkumar95