Wispr Flow, the Bay Area voice-dictation startup founded by Tanay Kothari, has raised $280 million at a $2 billion valuation, with India now its second-largest market by both users and subscribers. The company has hired a dedicated India operations lead, launched on Android, and put real product effort into a Hinglish voice model - the version of the product built for how people in India actually speak, not the version translated after the fact from English.
That last detail is the story most funding coverage will skip past on the way to the valuation number, and it’s the one worth sitting with.
Most companies treat India as a language problem to solve later
The standard global product playbook ships in English, waits for organic adoption, and treats “localisation” as a backlog item to revisit once a market proves itself. It’s a sequencing choice that quietly assumes the local-language version of a product is a lesser, later concern - something bolted on, not built in. For text products that assumption is survivable. For a voice product, it’s close to fatal, because voice has nowhere to hide an English-first design. A user either speaks naturally into the product or they perform a cleaner version of themselves for it, and most people won’t keep doing the second one for long.
Hinglish isn’t a compromise version of the product
Wispr Flow’s bet is narrower and more specific than “support more languages.” Indian professionals overwhelmingly code-switch - starting a sentence in English and finishing it in Hindi, or the reverse, sometimes mid-word. A voice model trained to expect one language at a time fails constantly against that pattern, and the failure reads to the user as the product not understanding them, not as a language-coverage gap. Building for code-switching from the start, rather than treating it as an edge case to patch later, is what actually made the product usable for a market most competitors were still describing as “English plus a roadmap item.”
What the rest of the market should take from it
The lesson here isn’t really about voice AI. It’s about the gap between products that are available in India and products that are built for it. Availability is a translation. Fit is a design decision made early enough that it shapes the product instead of trailing it. Wispr Flow becoming the company’s fastest-growing market wasn’t a coincidence of timing - it followed a decision to treat the way Indians actually talk as the primary design constraint, not a localisation task queued behind the “real” product.
That’s a more expensive way to build. It’s also, on the evidence of where the growth actually showed up, the only way that reliably works.