It was not so much a coincidence as a statement that the first major book NAAV AI translated was Vikram Sampath’s My Name is Gauhar Jaan into Kannada. Published in English in 2010, the award-winning biography had never reached Kannada-speaking readers, even though Sampath grew up in Bengaluru and Gauhar Jaan spent her final years in Karnataka.
For Sampath, the choice was symbolic. More than a century after Gauhar Jaan became the first Indian artiste to commercially record her voice on the gramophone, her story became one of the first full-length books to be translated into Kannada using AI-assisted technology.
“As a historian, I look at patterns,” says Sampath. “When Gauhar Jaan embraced recording technology, many musicians said it would destroy music or anger the gods. Today, writers and publishers have similar doubts about AI. History is repeating itself.”
The brainchild of Sampath and technologist Sandeep Singh Chauhan, Bengaluru-based NAAV AI—short for Navigating AI Across Vocabularies, and a play on the Hindi word naav (boat)—uses artificial intelligence and human expertise to translate long-form content such as books. It currently supports 11 Indian languages, including Hindi, Kannada, Tamil and Malayalam, as well as five foreign languages: French, German, Russian, Italian and Spanish.
The idea grew out of a frustration Sampath knew well as an author. “Only about five or six per cent of India consumes content in English. The actual soul of Bharat is in the Bharatiya bhashas,” he says. “But translation in India has largely remained a manual, mechanical, expensive and time-consuming process.”
Its flagship product, Translit, generates a first draft of an odd 300-page book in about an hour. It is then reviewed by language experts and professional translators, some from the Central Institute of Indian Languages Mysuru, who refine the text before publication. “The heavy lifting is done by AI, but the final product is human-refined,” adds Chauhan. “Every language has cultural and linguistic nuances that technology alone cannot fully capture.”
The result is a dramatic reduction in time. A 300-350 page book that would ordinarily take six to seven months can now be completed in around four weeks. “This isn’t about replacing translators,” says Chauhan. “Instead of translating one book in six months, the same translator can now work on six or seven books. Their role becomes more valuable.”
Winning over publishers, however, has required patience. Sampath says much of his time is spent explaining that NAAV is AI-assisted rather than AI-driven. “It is natural for people in the creative industry to have doubts,” he says. “The fear is that AI will write the book or replace the translator. But the human expert remains central to the pipeline.”
The other product ZuNAAV creates audiobooks using AI-assisted technology in no time. “We make books come alive through immersive background sound, special effects and music. Our audio engineers and language experts refine the final version before it reaches the publishers,” adds Chauhan. The company has also successfully tested a Live Translation App with the spiritual organisation of Sadguru Madhusudan Sai, expected to go live shortly.
The startup has received funding support from Ola founder Bhavish Aggarwal and investors such as Asha Jadeja Motwani, Prashanth Prakash and Subrata Mitra.
For Sampath, the larger ambition goes beyond technology. “If technology helps literature travel across languages,” he says, “India’s languages will grow together rather than compete with one another, as they most often do in current times.”