Some publicly available TTS datasets that can be used for training neural TTS methods are catalogued here
List of publicly available TTS datasets for English
Publicly available TTS datasets for Indian languages
The audio lab at IIT Madras has made publicly available studio quality datasets for 13 Indian languages in both genders, with an average duration of 10 hours per speaker. The database can be accessed online here.
The available languages are - Assamese, Bengali, Bodo, Gujarati, Hindi, Kannada, Malayalam, Manipuri, Marathi, Odia, Rajasthani, Tamil, Telugu