BHASHINI and the Indian Institute of Science have developed text-to-speech models for Sourashtra, according to a 16 October announcement from the electronics and IT ministry.
The release describes four versions: male and female voices for text written in Tamil script and in Sourashtra script. The system uses the VITS speech architecture and a speech collection of more than 28 hours assembled at IISc’s SPIRE laboratory.
The models are intended to support uses such as language learning and accessibility. The training corpus is available through AIKOSH. The announcement does not provide independent evidence of performance across all dialects, speakers or real-world settings.