loading...
Wispr Flow transcribes natural speech into writing that matches its destination — Slack sounds like Slack, clinical notes sound clinical. $81M raised, $2B valuation in talks.
ENTRY ANGLES
Build clinical dictation with structured EHR field mapping and missing-info alerts · Build vertical voice layer for legal document generation · Build field technician documentation tool with domain-specific vocabulary
VERTICALS
CAPABILITIES
HL7 FHIR expertise, EHR system integration, Clinical NLP, Regulatory compliance (HIPAA)
Mac, Windows, iOS, Android. Any application, no integrations required. Output calibrated to the writing register of the destination app. Three years after founding and $81 million across five funding rounds later, Wispr Flow is the first voice dictation product that professionals use as their primary input method rather than a fallback for long-form content.
Tanay Kothari and Sahaj Garg founded Wispr in 2021. The technical problem they set out to solve was not transcription accuracy — modern speech recognition has been reliable enough for years. It was register: spoken language and written language are different, and every tool that treated them identically produced text that needed more editing than the dictation saved time on. Flow's model learns the user's vocabulary, typical sentence structures, and recurring terms, then outputs clean prose calibrated to the context. A message dictated into Slack arrives formatted as a Slack message. An email drafted by voice lands as prose appropriate to email. Clinical notes, code comments, and formal documents each get a different output even from the same utterance.
The Pro plan is $15 per month and includes Command Mode — voice-driven editing that lets users rewrite, shorten, and redirect text already on screen. The free tier runs at 2,000 words per week. As of May 2026, Wispr is reportedly in talks with Menlo Ventures to raise $260 million at a $2 billion valuation.
The knowledge worker writing load has grown faster than typing speed. A senior employee in a messaging-heavy organization sends somewhere between 200 and 400 written messages per day across Slack, email, and documents. At that volume, the mechanical act of typing is the bottleneck — not the thinking. Voice should solve this, and for thirty years the tools existed to solve it. They failed because they produced output that read like dictation, which is a different register than the output the user needed.
Flow's per-app context model is the structural solution. By modeling the relationship between spoken input, the application receiving the output, and the user's individual writing patterns, Flow closes the register gap without requiring the user to consciously adapt their speech. The result is that professional users can dictate into any application and receive output they do not have to rewrite — which is the specific friction that broke every prior tool at scale.
A $2 billion valuation for a dictation app is not a bet on consumer voice input. It is the size of the claim that voice becomes the default input method for professional knowledge work — a market where per-seat pricing and enterprise compliance requirements produce meaningfully higher revenue per user than consumer subscriptions.
Wispr is building the general-purpose voice layer — any application, any context, any register. The opportunity for vertical specialists is to go deep where general-purpose breaks down.
Medical dictation is the most concrete example. A physician dictating a complex oncology note needs the model to recognize staging classifications, drug names with specific dosing conventions, and clinical abbreviations that general language models frequently hallucinate. The medical dictation market is approximately $1.5 billion, growing with EHR adoption, and currently dominated by Nuance's DAX product at $150+ per physician per month — a price point that most hospital systems treat as expensive but unavoidable.
The gap is not transcription. It is structured documentation: a system that listens to a physician-patient encounter, transcribes it, maps the output to the structured fields in the relevant EHR, flags missing required information, and alerts the physician before they leave the room. That product does not yet exist as a third-party offering compatible with multiple EHR systems. Building it requires HL7 FHIR expertise, EHR-specific field mapping, and clinical workflow knowledge — which is precisely why DAX has held the position for three years without a credible challenger despite pricing that leaves obvious room for competition.