Summary: The two-year wait for Wispr Flow on Android ended on February 23, 2026. That route carries permission friction: before dictation works, users must flip an accessibility toggle in system settings — precisely the kind of switch security-conscious people are trained to distrust, a genuine adoption hurdle for a consumer app. The AI dictation app built for Mac and Windows first, added iOS in June 2025, and left Android users at the back of the line.
The two-year wait for Wispr Flow on Android ended on February 23, 2026. The AI dictation app built for Mac and Windows first, added iOS in June 2025, and left Android users at the back of the line. In the company’s blog, CEO Tanay Kothari made the reasoning plain: “We were not going to ship Flow on Android until it was genuinely unmatched.” During a limited window, the Android app gives dictation away free and unlimited.
Wispr frames the stakes with big numbers: Android counts roughly 4 billion daily users, and the company says hundreds of thousands of them sat on its waitlist. As Kothari tells it, “voice on Android feels like a primary interface, not a fallback feature” — for the first time. The shipping cadence is its own statement: in this category, Wispr measures itself against an experience bar rather than a market-share race.
The Floating Bubble: Dictation Without a Keyboard Swap
The entire interface on Android is one bubble hovering above whatever app you’re in: press and hold to speak, let go to finish, or tap once to start and tap again to close. Where the iOS version ships a dedicated keyboard, this build runs systemwide on top of Android’s Accessibility Service. That route carries permission friction: before dictation works, users must flip an accessibility toggle in system settings — precisely the kind of switch security-conscious people are trained to distrust, a genuine adoption hurdle for a consumer app. The app cleans filler words out of your speech and formats the text to match the app context, and editing behaves identically on Android, iPhone, Mac, and Windows, so your workflow survives a device switch. In Kothari’s words: “If you can speak, you can now out-type your keyboard.”
A Ground-Up Infrastructure Rewrite
A port alone was never the plan. Alongside the launch, Wispr finished a complete infrastructure rewrite that leaves Flow 30% faster on every platform, and the reasoning is arithmetic: multiply latency by billions of daily uses and you get an enormous pile of wasted user time. Official claims put the app at 3x the accuracy of competing voice models, while Kothari judges himself by a harsher measure: “Even a 1% error rate is 100x too high.”
One Hundred-Plus Languages, and a First Hinglish Model
Translation spans more than 100 languages, and this release adds a Hinglish model for speech that mixes Hindi and English. Wispr calls it the first voice model to transcribe Hinglish directly instead of routing it through Devanagari script first. Kothari described it as “one of those times when I just had to build something for me”: among family and colleagues back home, English and Hindi weave together on their own. The demand is visible — early-access users dictated more than 1.3 million English words over recent days.
The Money and the Competition
Funding sits at $81 million in total: $30 million led by Menlo Ventures in June 2025, then $25 million led by Notable Capital in November 2025, with sources cited by TechCrunch putting the last round’s valuation around $700 million. The field is already crowded: Typeless reached Android a month earlier, and Google’s Gboard voice typing is the preinstalled baseline. The structural divide is focus — Wispr plays dictation specialist, tuning accuracy and latency, while Gboard keeps voice as one general-purpose feature among many. On the Mac side, users can already add custom words and standing dictation instructions, and carrying that personalization to mobile is the obvious next step. Wispr’s long-run proof sits in desktop data: after six months, users reportedly type less than a third of their text manually, which is the basis for the company’s claim that it has “finally displaced the 150-year-old keyboard.”
Three Lessons for Product Teams
Lesson one: platform openness can itself be the feature. Android “finally gave us the freedom to build the voice experience we always wanted,” Kothari said — a systemwide floating overlay simply isn’t buildable on more restrictive platforms. Lesson two: retention is built on cross-platform consistency; formatting and editing behavior must match everywhere before people treat voice as their default input. Lesson three: an AI app beating a built-in system feature is rare, and this is one of the few — out-shipping a preinstalled voice keyboard takes hard metrics like accuracy and latency, not a longer feature list.