Skip to content

Dataset · ai4bharat

IndicVoices-R: Multilingual Speech Corpus

ai4bharat/indicvoices_r

Large audio dataset with transcripts in 22 Indian languages for training text-to-speech systems.

What's inside

Contains paired audio recordings and texts with a total duration of 1,704 hours from 10,496 speakers across 22 Indian languages (including Hindi, Bengali, Gujarati, and others). Sample size ranges from 100k to 1M recordings. Data was derived by adapting an existing automatic speech recognition corpus for text-to-speech tasks.

What you can do with it

  • Train multilingual and multi-speaker TTS models for Indian languages
  • Fine-tune natural speech generation systems considering speaker demographics
Tasks
Text → speech
Languages
Assamese, Bangla, Gujarati, Hindi, Kannada, Kashmiri, Malayalam, Marathi and 4 more
Size
100K–1M rows
Downloads this month
15.6K
Likes
56
Access
on request on Hugging Face

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("ai4bharat/indicvoices_r")