Canto: Next-Gen AI Dictation
Wispr AI Secures $280M Series B, Launches Canto to Revolutionize Real-World AI Dictation
Key Points
Wispr Flow has processed over 60 billion words and is used by more than 10,000 enterprises, with nearly all Fortune 500 companies relying on its platform.
Wispr's proprietary model Canto dramatically improves performance in real-world noise, cutting error rates from over 30% to about 5–10% when integrated with Flow.
The competitive landscape for AI dictation is expanding beyond Otter and Granola to include players like Willow, Monologue, Aqua and Superwhisper, signaling heightened competition.
Wispr is partnering with hardware makers, such as the Oasis ring, to enable dictation on devices without having to speak loudly.
Wispr has launched Wispr Interface Labs, led by Ariya Rastrow, to explore new interfaces and product directions beyond traditional dictation.
Wispr AI has raised $280 million in a Series B round led by Menlo Ventures, valuing the company at $2 billion, according to Silicon Angle. The deal brings Wispr's total funding to $361 million and cements its place as one of the fastest-growing AI voice startups in the market.
Alongside the funding, Wispr unveiled Canto, its first proprietary speech model. Canto is built to handle real-world noise — wind, crowds, interruptions — cutting transcription error rates from over 30% down to just 5–10%, The AI Insider reported.
Wispr Flow's Staggering Enterprise Scale
Wispr Flow has already processed more than 60 billion words. Over 10,000 enterprises use the platform, including nearly every Fortune 500 company, according to Tech Startups. That kind of adoption is rare for a startup at this stage.
Flow works across apps and operating systems as a dictation layer. Users speak, and Flow types for them — in emails, documents, Slack, and more. The goal is a zero-edit workflow, where what you say is exactly what you meant to write.
Canto Cuts Errors in Noisy Real-World Settings
Most speech models are trained on clean, quiet audio. Canto is different. Wispr trained it on real-world conditions — background chatter, wind, and overlapping voices. The result: error rates fall from roughly 30% to between 5% and 10%, The AI Insider reported.
That gap matters enormously outside a quiet office. A 30% error rate means roughly one in three words is wrong. At 5–10%, transcription becomes genuinely usable in the real world — on a commute, in a meeting room, or at a noisy cafe.
Wispr Expands Into Meetings and New Hardware
Wispr is moving beyond pure dictation. The company launched Notetaker, a tool that joins meetings, records them, and generates summaries. It is also partnering with hardware makers — including the Oasis smart ring — to let users dictate without speaking loudly, Yahoo Finance reported.
Wispr has also launched Wispr Interface Labs, led by researcher Ariya Rastrow, to explore entirely new interfaces and product directions. The lab signals that Wispr sees dictation as just the starting point, not the finish line.
A Crowded Race for AI Voice Dominance
Wispr is not alone. Otter.ai and Granola are well-known competitors, but newer players are entering fast. Willow, Monologue, Aqua, and Superwhisper are all vying for the same market, according to 963 Jack FM. The space is getting crowded quickly.
What sets Wispr apart, at least for now, is scale. Sixty billion words processed and near-universal Fortune 500 adoption are hard to match. The $280 million raise gives Wispr the runway to build faster than rivals and lock in enterprise customers before the competition catches up.
Sentiment
Mostly PositivePublishers
13
Articles
7
Reach
20
