IISc Bengaluru and BHASHINI launch text‑to‑speech models for Sourashtra language

What happened

The Digital India BHASHINI Division, a unit of the Digital India Corporation under the Ministry of Electronics and Information Technology, announced that it has partnered with the Indian Institute of Science (IISc), Bengaluru, to release four text‑to‑speech (TTS) models for the Sourashtra language. The models, built on the VITS (Variational Inference with Adversarial Learning for end‑to‑end Text‑to‑Speech) architecture, can synthesize speech directly from characters written in either Tamil or Sourashtra script. Each script is available in a male and a female voice, giving users a choice that mirrors natural speech patterns.

How the models were built

The speech data were collected at IISc’s SPIRE Lab. Researchers compiled a dedicated Sourashtra corpus that contains more than 28 hours of gender‑balanced recordings – roughly 10,000 utterances from a male voice artist and 11,400 from a female voice artist. The recordings include both purpose‑written sentences and excerpts from literature, each transcribed in both scripts. The audio is generated at a sampling rate of 22,050 Hz and the system works at the character level, removing the need for a separate phonemiser.

Why it matters

Sourashtra, spoken by around 2.48 lakh people according to the 2011 Census, has traditionally had a sparse digital footprint. By providing high‑quality synthetic voices, the new models open the door to a range of applications: reading e‑books aloud, powering language‑learning apps, enhancing accessibility for visually impaired users, and feeding downstream research in low‑resource speech synthesis. The models are already uploaded to BHASHINI’s AIKOSH platform, where developers can access the SPICOR TTS 3.0 Sourashtra Corpus and integrate the voices into public‑facing services.

Part of a national push

The launch aligns with the government’s Digital India agenda, which seeks to make digital services available in every Indian language. BHASHINI currently supports 36 Indian text languages, 23 voice languages and 35 international languages, and it aims to embed AI‑driven language capabilities across government portals, education portals and citizen‑centric apps. Adding Sourashtra to this ecosystem demonstrates the commitment to linguistic diversity and showcases India’s growing expertise in AI‑enabled multilingual technologies.

Looking ahead

With the corpus now public on AIKOSH, academic teams and private developers are encouraged to build on the foundation. Potential extensions include speech‑to‑text models, real‑time translation, and voice‑enabled assistants that understand Sourashtra. The collaboration between a premier research institution and a national digital initiative signals a scalable model for bringing other under‑represented languages into the digital mainstream.