During his recent visit to India, Microsoft executive chairman and CEO Satya Nadella said he was impressed with the work “Bhashini” was doing, which resulted in at least 10 people reaching out to Shiva Kumar HR, the CEO of AI solutions provider Bhashini, only to realise they have dialled the other ‘Bhashini’.
Nadella was highlighting the work done by Bhashini, an initiative by the Indian government which leverages the power of artificial intelligence (AI) and other emerging technologies with an aim to foster content in regional languages and enable easier internet access for all citizens.
Shiva Kumar’s Bhashini is not very different. The Bengaluru-based startup offers a text-to-speech (TTS) software that enterprises use to produce human-like interactive voices unique to their organisations.
“Like every company has a different website, they are also increasingly opting for software which helps them develop an AI-generated voice which, when heard, can be associated only with their organisation,” Kumar explained.
Explaining his company’s offerings further, just like Amazon’s Alexa could speak like Amitabh Bachchan, the same way Bhashini codes/programmes and makes the software listen to the original voice once. After the tool has learnt and captured the voice, any text could be fed as an input and the software would speak the voice of the original person.
Also read: Govt plan to repower old wind power plants could attract Rs 40,000 crore investments
Tapping into those offerings, TV channels were using Bhashini’s technology to air certain shows in their anchor’s voice and “viewers can never tell when their favourite anchor was speaking or when it was our software at play”, Kumar said.
Similarly, universities were leveraging Bhashini’s technology to feed audio books into their database. Visually impaired students use the software to listen to the content in a language of their choice and a voice they felt most comfortable in, which would be as good as a human’s, Kumar added.
Currently, Bhashini offers around seven languages and Kumar said his 12-member team was working hard to include more regional languages. But “larger enterprises find it difficult to trust a small startup like ours, so they instead opt for Microsoft’s Nuance or Google’s AI offerings. That has been a challenge for us”, said Bhashini’s Kumar, who is also a former IBM employee.
“I have worked with Big Tech firms and know that Bhashini’s homegrown software is leaps and bounds ahead of all, else I wouldn’t leave my well-paying job and work on the software since 2008. We rolled out the services only in 2020 – took 12 years to perfect it,” he added.
Kumar, a PhD from the Indian Institute of Science (IISc) in Bengaluru, said his company wasn’t too big, the revenue was a few million but was profitable, partly helped by the grants it has received from the state government.
“While I would be interested to collaborate with enterprises, I am not keen on venture capital because the money comes with a lot of caveats,” Kumar said.
Also read: Mandatory joint audit to start with big firms
Bhashini’s next software rollout would allow deep dubbing, especially in movies. Global hits like Kantara and Bahubali would attract a much wider viewer base if the films released in Russian, for instance, Kumar believes. “We want to also build technology that can make characters speak in any global language, in their own voice, not a native speaker dubbing it. That way the original essence would remain but that is still a few years away,” he said.
Content in native languages has seen an uptick of late with even edtech companies branching out to teach students in a language of their choice, as reported earlier. A survey had concluded that 53% of Indians who don’t access the internet have said they would begin using the web if it had content available in their native languages.
That was also a key reason why the Indian government rolled out its own project, Bhashini, which empowers citizens to speak in their own language while talking to speakers of other Indian languages, by using AI and natural language processing (NLP) tools.
