English remains a barrier between many Indian readers and the country’s literary works. Vikram Sampath, historian, author and Co-founder of Naav AI, wants to use AI to change that by making books available in Indian languages without losing the author’s voice.
Sampath has spent two decades writing books, but getting them translated into Indian languages has often been a lengthy process.
Some translations take years to reach readers, while others never get published because of translation costs, shortage of skilled translators, and gaps in the publishing industry’s distribution network.
Sampath sees AI as a way to make literature available to a wider audience across languages.
“Only 5 to 6% of India, I think, consumes content in English. The real Bharat consumes content in Bharatiya Bhashas like Hindi, Bengali or Marathi,” says Sampath, who authored books including Savarkar, Bravehearts of Bharat and Tipu Sultan.
Founded in January 2025, Naav AI, short for “Navigating AI Across Vocabularies”, uses agentic AI frameworks to translate long-form content across languages and convert text into other formats, including audio and video.
The idea behind Naav AI emerged from a conversation between Sampath and his friend and now co-founder Sandeep Singh Chauhan about the challenges of translating books into Indian languages.
The idea was to use AI to help Indian publishers meet the demand for content in regional languages and take Indian stories to audiences abroad.
Naav AI’s publishing clients include BluOne Ink, and the startup is working with the Foundation for Indian Historical and Cultural Research (FIHCR) to translate children’s books on Indian history
The company has received 200,000–250,000 in early-stage funding from individual investors, including Ola founder Bhavish Aggarwal, Asha Jadeja Motwani, Prashanth Prakash and Subrata Mitra.
Its flagship product, TransLit, uses AI-assisted translation and human review to localise books and other long-form content, while ZuNAAV-FM is being developed as an audiobook and audio-content platform.
Translate Books in Hours
Naav AI uses a multi-model, agentic workflow to generate the first draft of a translation. Sarvam AI is its primary technology partner, while the startup also uses models from Google and OpenAI.
Sampath says initial translations into Indian languages are typically 80% to 85% accurate, requiring human intervention before they are ready for publication.
“As a creative person and an author myself, I would have been very sceptical if the output was from a machine language,” he says.
The startup works with language experts, including specialists from the Central Institute of Indian Languages in Mysuru, to refine translations and preserve the original text’s style, idioms and cultural context.
Human editors ensure the translated text retains the nuances of the original work.
Naav AI currently supports 11 Indian languages, including Hindi, Marathi, Kannada, Tamil, Telugu, Malayalam, Odia, Bengali, Gujarati, Assamese and Punjabi. It also works with German, Russian, French, Spanish and Italian.
Sampath says the startup’s translation tool can translate a 400-page book in about an hour. The subsequent human review and refinement process takes longer, with the full workflow reducing the time to produce a publish-ready translation to three to four weeks.
A typical 300-page book could otherwise take around six months to translate, he notes.
The startup’s agents also maintain context across lengthy manuscripts, tracking characters, names and terminology throughout a book. This helps ensure that a character introduced in an early chapter is referenced consistently later in the story.
As human translators make corrections, the system learns the preferred style and applies those changes to subsequent sections, helping improve consistency.
For languages such as Hindi, where training data is more readily available, Sampath says Naav AI is targeting accuracy levels of up to 90%.
AI Could Increase Translators’ Earnings
Sampath believes AI-assisted translation need not eliminate jobs in the publishing industry. Instead, it could allow translators to work on more books and increase their earnings.
A translator who previously spent one or two years translating a single book could potentially work on six or seven books over the same period with AI assistance, he says.
The startup credits human experts for their contribution to the final product. Translated books carry the name of the language specialist as an ‘Anuvad Sahayak’, or translation assistant, along with Naav AI’s logo, depending on the publisher’s preferences.
The publisher retains the copyright, while Naav AI operates as a business-to-business service provider. Sampath says the company has begun working with several Indian publishers and some multinational publishing houses.
The company has not adopted a subscription model for long-form translation because it requires human intervention, he adds.
Audiobooks
Beyond text translation, Naav AI is developing audiobook production through a product called ZuNAAV-FM.
Sampath believes audio can help literature reach people who do not regularly read physical books, including those who may be unable to read. He also highlights the growing consumption of audio and video content, particularly among urban audiences who listen to content while commuting.
Through ZuNAAV-FM, the startup uses agentic workflows to orchestrate audiobook production. The process begins with a mood board that maps the book’s characters, including their genders, ages and accents. Producers can then incorporate background music and sound effects to create an immersive listening experience.
Sampath cites his book on Tipu Sultan as an example. A battle scene could feature trumpets, rockets, elephants trumpeting and horses neighing to recreate the atmosphere for listeners.
The startup sees opportunities to supply high-quality audio content to platforms such as Kuku FM, Pocket FM, Audible and Spotify.
AI Can Help Indian Languages Thrive
For Sampath, Naav AI is driven by both creative and commercial considerations. Translating his work allows him to reach readers who may not understand English, while the business model creates an opportunity to serve a larger market.
He also sees language accessibility as an ethical concern, arguing that access to quality literature and knowledge should not depend on English proficiency.
India has 22 scheduled languages and thousands of dialects, he says. Technology could help make content accessible across these linguistic communities while also taking Indian stories to international audiences.
“Technology can help unite where political divisions among languages actually divide,” Sampath says.
He argues that government recognition and funding alone cannot ensure a language’s survival. Languages need a steady supply of high-quality content across formats, including books, audio, video and short-form media.
AI-Generated Writing
Despite his support for AI-assisted translation, Sampath believes there should be a distinction between using AI to translate existing work and using it to create original literature.
He says AI tools could help writers structure sentences, sharpen prose and edit content, much like a conventional copy editor. However, he believes the author should remain responsible for the original creative work.
“We are not creating original content in a particular language. We only help with translation,” he says.
For Sampath, ethical AI use in creative industries means keeping humans in control. AI can reduce the time and cost of translation and content production, but human expertise is still needed to preserve meaning, creativity and cultural context.