Wispr raises $280M to power up natural speech-to-text using AI

Wispr AI has pulled in $280 million in Series B funding at a $2 billion valuation, a big vote of confidence in a product category that still feels oddly underrated. Menlo Ventures led the round, with existing backers including Notable Capital, NEA, Neo Ventures, 8VC and MVP Ventures. New investors Acrew, Forerunner, Goodwater, Peak XV, Together Fund and PLUS Capital also joined in, lifting Wispr’s total funding to $361 million.

The company’s main product is Flow, a dictation app that plugs into other applications and lets people speak into any text field. It doesn’t just transcribe; it cleans up what you said, smoothing out grammar slips, filler words like “ums” and “ahs,” and half-finished thoughts so the result reads like an email, memo or document instead of raw speech. That makes it useful on mobile and desktop, and anywhere else typing feels like the slower option.

The more interesting piece is Canto, Wispr’s first proprietary AI model, which the company is previewing alongside the funding. Wispr says Canto was trained for messy, real-world audio: loud rooms, interruptions, dogs barking, overlapping voices and all the other background noise that breaks cleaner voice models. Most systems, the company says, are trained on very clean voice data. Canto was built to separate human speech from everything else.

If Wispr’s numbers hold up, that matters. The company says Flow’s error rate in noisy environments drops from 30% to nearly 5% to 10% when Canto is built in. That is the difference between a tool that works in theory and one people might actually use in a car, near traffic, or in an office without babysitting every sentence.

Wispr says people have written more than 60 billion words on Flow, and that the product has been used by almost all Fortune 500 companies and over 10,000 enterprises. It also has a well-known user in Domantas Sabonis, who says he uses Flow every day and switches among English, Spanish and Lithuanian. The bigger story, though, is that speech-to-text is moving from neat demo to everyday utility — and the noisy world is finally the test.