Context This dataset is a preprocessed for Tacontron2 model version of Mozilla Common Voice dataset https://voice.mozilla.org/en. Only Indian accent is considered (marked 'indian'). The dataset consists of 9870 utterances, 2870293 mel frames, 572865836 audio timesteps (9.95 hours). Max input length (text chars): 122 Max mel frames length: 887 Max audio timesteps length: 177280 Content "audio" folder consists of three subfolders (train, test, dev). Each of them includes folders, named by the speaker ID. Speaker ID folder has: - "audio.mp3