Pariksha SaathiBHASHINI and IISc Bengaluru build text-to-speech models for Sourashtra, a low-resource language of Tamil Nadu
The Digital India BHASHINI Division of MeitY, with the Indian Institute of Science (IISc), Bengaluru, has developed text-to-speech (TTS) models for Sourashtra, a minority Indo-Aryan language spoken mainly by the Sourashtra community in Tamil Nadu and by about 2.48 lakh people according to the 2011 Census, PIB reported on 16 October 2026. The four models, built on the VITS (Variational Inference with adversarial learning for end-to-end Text-to-Speech) architecture, offer a male and a female voice, each for text in Tamil script and in Sourashtra script, generate audio at 22,050 Hz and work directly on character-level text without a separate phonemizer. They were trained on a gender-balanced Sourashtra speech corpus of more than 28 hours recorded at IISc's SPIRE Lab (about 10,000 utterances for the male voice and 11,400 for the female voice), now published as the SPICOR TTS 3.0 Sourashtra Corpus on BHASHINI's AIKOSH platform. The models are meant for reading Sourashtra text aloud, language learning, accessibility tools and research on low-resource speech synthesis.
Drafted with AI assistance from the official PIB release linked above and checked against it by a second AI pass. Not line-edited by a human.
Learning context
The Census of India 2011 recorded 19,569 raw mother tongues, grouped into 121 languages: the 22 Eighth Schedule languages and 99 non-scheduled languages; Sourashtra is among the non-scheduled ones. The Sourashtra community traces its origin to the Saurashtra region of Gujarat and has lived for centuries in Tamil Nadu, especially around Madurai. Text-to-speech systems convert written text into spoken audio; a low-resource language is one with little digital text and speech data to train such systems.
Reference: Census of India 2011 language data
Check yourself: is Sourashtra written in only one script?
No. The release provides each voice for text in both Tamil script and Sourashtra script, four models in all, because the language is written in both.
Revision task: from this release, note who built the models, the corpus size (28-plus hours), the speaker count from the Census and the platform (AIKOSH), and keep them apart from the standing facts about the Census and the community above.
Test yourself
Sourashtra, for which BHASHINI and IISc Bengaluru released text-to-speech models on 16 October 2026, is spoken mainly in which state?
Show the answer
Answer: Tamil Nadu
Sourashtra is a minority Indo-Aryan language spoken mainly by the Sourashtra community in Tamil Nadu, by about 2.48 lakh people according to the 2011 Census; the models cover both Tamil and Sourashtra scripts.
Read the daily collection → · Editorial standards & corrections