India to develop voice-based AI models for greater accessibility

The government is prioritising the development of homegrown voice-based AI models over text-driven ones to enhance accessibility for a larger population. This initiative aims to enable users to interact with AI in their native languages through speech rather than text. Unlike global AI models such as OpenAI’s ChatGPT and Google’s Gemini, which primarily rely on text, India’s focus is on vocal interactions.

While major tech firms have introduced voice capabilities, their models are largely optimised for English and a few other languages. A key factor driving this shift is that many Indians are more comfortable speaking their native language than typing it. Abhishek Singh, additional secretary at the Ministry of Electronics and IT (MeitY), emphasised the market potential of such models.

With nearly 900 million active Internet users and around 500 million still offline, voice AI could significantly improve digital inclusion. Rural communities, farmers, and non-tech-savvy individuals could access services more easily through voice commands. For instance, a farmer could ask about crop diseases or weather in his dialect instead of navigating text-based interfaces. Similarly, voice AI could assist in healthcare, education, and government services.

Global AI firms have struggled to cater to India’s linguistic diversity due to limited training datasets. To address this, the government has received 67 proposals from startups, academia, and enterprises for developing foundational AI models. Of these, 22 focus on large language models, while 45 target domain-specific AI for sectors like healthcare and agriculture. A technical committee is reviewing these proposals, with selected projects receiving government funding.

Additionally, a national dataset platform is being developed to support AI training. By investing in voice AI, India aims to bridge the digital divide and establish itself as a leader in AI innovation.