The Digital India BHASHINI Division under the Ministry of Electronics and Information Technology, in partnership with the Indian Institute of Science Bengaluru, has created text-to-speech models for Sourashtra, a minority Indo-Aryan language. This technology allows computers to read Sourashtra text aloud in both male and female voices, addressing the digital gap for a language with limited online resources.
Key facts
- Four text-to-speech models developed: two speakers (male and female), each available in Tamil and Sourashtra scripts
- Speech corpus contains over 28 hours of gender-balanced recordings
- Male voice trained on approximately 10,000 utterances; female voice on approximately 11,400 utterances
- Sourashtra spoken by around 2.48 lakh people according to the 2011 Census, predominantly in Tamil Nadu
- Models use VITS architecture and generate audio at 22,050 Hz
- Speech corpus and models available on AIKOSH, BHASHINI’s platform for language resources
- Models operate directly on character-level text without requiring separate phonemizer
How the technology works
The models use a technique called VITS (Variational Inference with adversarial learning for end-to-end Text-to-Speech), which converts written Sourashtra text directly into spoken audio. The system was trained using recordings from male and female voice artists at the SPIRE Lab at IISc Bengaluru. These recordings included both composed sentences and passages from Sourashtra literature, ensuring the model learns natural and authentic speech patterns. Since the models work at the character level, they do not need an intermediate step to convert letters into pronunciations, making them simpler and more direct.
Why this matters for Sourashtra
Sourashtra is an Indo-Aryan language with relatively few digital resources compared to major Indian languages. Most speech technology and online content exists in languages like Hindi, English, Tamil and Telugu. By creating text-to-speech models, the technology brings Sourashtra speakers into the digital ecosystem, enabling them to access and interact with their language online. The dual-script capability is important because Sourashtra speakers use both the Sourashtra script and Tamil script, depending on context and preference.
Practical applications
The technology can be used in several ways. Educational institutions can develop language learning tools that read Sourashtra text aloud, helping students and adults learn the language more effectively. Digital accessibility improves for Sourashtra speakers who prefer or need audio content. Libraries and archives can preserve Sourashtra literature and make it accessible through audio. Government and public digital services can integrate Sourashtra speech capabilities. Researchers studying low-resource languages and speech synthesis have new data and models to build upon. The technology also creates a foundation for future applications in Sourashtra language technology.
AIKOSH platform and broader BHASHINI work
The speech corpus and models have been made available on AIKOSH, a platform run by Digital India BHASHINI Division that provides language datasets and resources for building Indian language technologies. This open availability means developers and researchers outside the original teams can use the resources to create new applications. Digital India BHASHINI, operating under the Ministry of Electronics and Information Technology, currently supports 36 Indian text languages, 23 Indian voice languages and 35 international languages. It aims to make government and public digital services accessible across India’s linguistic diversity through technologies including translation, speech recognition, speech synthesis and transliteration.
What this means for you
If you speak Sourashtra or work with the Sourashtra-speaking community, this technology enables new ways to access digital content and services in your language. Students can use speech-enabled learning tools. People who prefer audio content or have visual accessibility needs can now access Sourashtra text through speech. If you are a developer or researcher, the open availability of the corpus and models on AIKOSH provides a starting point for building new applications. For language preservation efforts, this technology helps ensure Sourashtra remains relevant and accessible in the digital age.