TrueFan AI

Here is a fact about celebrities that is obvious once you say it out loud but was apparently not obvious enough to stop several startups from building businesses on top of it: celebrities have exactly as many minutes in a day as everyone else. This is the constraint that TrueFan AI ran into in 2020, and it is, in a roundabout way, the reason the company exists in its current form.

The founders - Nimish Goel, Devender Bindal and Nevaid Aggarwal - started with a pleasant idea. Fans want to feel close to their idols. Goel had felt it himself, meeting his sports heroes. So the plan was a platform where a fan could pay for a personalized video greeting from a movie star. This is a real business; other companies do it. The problem is that it does not scale, because the input that matters - the star's time - is the one thing you cannot manufacture. A celebrity can record only so many "Happy Birthday, Rohan" clips before the whole enterprise becomes a scheduling nightmare.

Most companies, faced with this, would grind. TrueFan did something more interesting: it treated the bottleneck as a technical problem. If the expensive, scarce input is the celebrity's performance, then the obvious move is to learn the performance once and generate the rest. In late 2021 the company launched its first AI avatar, of the actor Ranveer Singh. Over the next couple of years it built avatars for more than 150 celebrities - Shah Rukh Khan, Ranbir Kapoor, Kareena Kapoor - each one a licensed digital stand-in that could say things the human never sat down to record.

The pivot nobody roots for

Now, if you have built AI avatars of the biggest names in Bollywood, the temptation is to stay in that world forever. It is glamorous. It gets press. It is also, it turns out, not where the money is. In 2024 TrueFan did the unglamorous thing and pivoted from a consumer celebrity app to a business-to-business video platform. The celebrities didn't disappear - they became a feature, a way for a brand to run an ad with an AI-generated star. But the core product became something far more boring and far more lucrative: infrastructure for enterprises that need to make an enormous amount of video and don't want to book studios to do it.

The pitch to enterprises is essentially an arithmetic argument, and arithmetic arguments are the ones that close. A bank shoots one five-minute video with a spokesperson. TrueFan's models take that single recording and generate personalized, localized versions - different names, different languages, different details - at a rate the company puts at up to 500,000 videos per minute, across more than 175 languages. The lip-sync moves, the voice is synthesized, the gestures track. One shoot becomes a fleet.

Why 175 languages is the whole point

It is easy to read "175 languages" as marketing garnish. In India it is closer to the entire thesis. A customer in Chennai and a customer in Kolkata may not share a first language, and a bank that wants to speak to both in their own tongue historically faced a translation-and-reshoot problem that scaled linearly with ambition. TrueFan turns that into something closer to a dropdown menu. The company's models handle the facial dynamics, gestures and voice synthesis so the output reads as a person, not a caption track.

The customer list is the kind that makes investors comfortable, because it is not a list of other startups. It is HDFC Bank, Bajaj Finance, Zomato, Cipla, BharatPe, Axis Max Life Insurance, Goibibo - banks and insurers and consumer brands, more than 100 of them, the sort of buyers who pay real money on annual contracts and don't churn on a whim. The company says it has delivered more than 10 million AI videos.

What the machine is actually doing

It is worth being precise about what "AI video" means here, because the phrase is doing a lot of work in a lot of pitch decks that don't deserve it. TrueFan's system isn't stitching together stock clips or slapping a synthetic voice over a slideshow. The company built its own video foundation model, trained to reproduce three things that make a talking human read as human: facial dynamics, gestures and voice. Feed it one recording and it learns the person well enough to put new words in their mouth - the lips move to match a script the speaker never read, in a language they may not speak, and the head and hands move the way a person's do when they talk. The output is meant to survive the scrutiny of a customer watching it on a phone, which is a higher bar than a demo reel.

The volume claim - up to 500,000 videos a minute - is really a claim about architecture rather than showmanship. Personalization at that scale only works if the expensive part (learning the performance) happens once and the cheap part (swapping in a name, a number, a language, a product detail) happens on demand. That is the difference between a novelty and infrastructure, and it is the difference enterprises are paying for. A marketing team can send a genuinely different video to every customer in a database and treat the whole thing as a rendering job.

The team and the discipline

TrueFan runs on a team of roughly 100 people, and the more interesting thing about it is not the headcount but the temperament. Building a proprietary video model is an expensive, unglamorous, research-heavy slog, and it is the kind of work that companies chasing quarterly celebrity headlines usually don't have the patience for. The tell is the 2024 pivot itself: walking away from a recognizable consumer brand to sell rendering infrastructure to banks is a decision that requires a founding team willing to be less famous in exchange for being more valuable. Not every startup has that trade in it.

The field it is racing in

TrueFan is not alone. Synthesia and HeyGen are the best-known names in generative avatar video, both well funded and global; D-ID and Colossyan crowd the same space. In a straight feature fight, none of these companies wins on a single axis forever - the models converge, the demos start to look alike. What TrueFan has is a market shape that plays to its strengths: India, where the multilingual, high-volume, cost-sensitive use case is not an edge case but the default, and where it already has enterprise relationships that a foreign entrant would spend years building. The Series A money aimed at Southeast Asia, the Middle East and the US is a bet that this same shape - many languages, price-sensitive scale, enterprises that need to talk to everyone at once - travels.

The money

In June 2026 TrueFan raised a $10 million Series A led by Baring Private Equity Partners India and Z3Partners, with IAN Alpha Fund and 3Lines Venture Capital participating, closing at a post-money valuation of $40 million. That follows a $4.3 million seed round back in 2020, when the company still thought it was in the fan-engagement business. FY25 revenue came in around Rs 17.1 crore - roughly $2 million - up 131% year on year, which is the kind of growth curve that tends to follow a pivot that actually worked. The new capital is earmarked for expansion into Southeast Asia, the Middle East and the United States, plus continued investment in the underlying models and real-time AI video agents.

There is a version of the AI-hype cycle where every company claims to be building the future of everything. TrueFan is a useful counterexample, because what it is actually doing is narrower and therefore more believable: it removed one expensive, repetitive step - the video shoot - from a workflow that a lot of large companies run constantly. The AWS Conclave named it Gen AI/ML Disruptor of the Year in 2025, which is a nice trophy, but the more telling signal is that banks keep signing up. Its competition - Synthesia, HeyGen, D-ID, Colossyan - is well funded and global, so the international expansion is not going to be a stroll. What TrueFan has is an India-first head start in exactly the multilingual, high-volume market where the arithmetic bites hardest.

The tidy irony is that a company founded to help fans get a little closer to the stars ended up building its business by making the stars optional. The technology that was supposed to scale a celebrity's warmth turned out to be more valuable scaling a bank's onboarding flow. That is not the story anyone pitches in a seed deck. It is, apparently, the one that raises the Series A.