IISc releases voice AI model for 65 Indian languages

The Indian Institute of Science (IISc) has released an open-source voice AI model named SraVaani. The model is designed for speech-to-text conversion, meaning it turns spoken audio into written text. It supports 65 Indian languages, and those languages span 10 scripts. The model files have been made available on Hugging Face, a platform used for hosting open-source AI models. The distribution is under the MIT licence, a permissive open-source licence.

As an open-source release, SraVaani can be downloaded and used by developers and researchers. The MIT licence places few restrictions on use, modification, and redistribution. Hugging Face provides a central repository where the model can be accessed. The report gives the number of supported languages and scripts as the main details. It does not specify which 65 languages are included, nor which 10 scripts are covered, and technical details such as model size, architecture, and training data are not mentioned.

The model's stated purpose is to convert speech in Indian languages into text. A model with support for 65 languages could be used in applications that require transcription of spoken content. The availability of the model on Hugging Face allows outside parties to incorporate it into their own projects. The report does not include performance benchmarks or comparisons with existing speech-to-text services. Additional documentation would be required to evaluate the model's accuracy and reliability.