Qenzo AI Dictation Keyboard vs Wispr Flow

The short answer

Qenzo and Wispr Flow both turn speech into finished text, but they run in different places. Qenzo is a released system keyboard on Android plus a native desktop app on macOS, Windows and Linux, with an iPhone keyboard (requires iOS 17+) built the same way. Wispr Flow is a Mac/Windows overlay with an iPhone keyboard and an Android floating bubble, and its own documentation lists Linux, iPad, Chromebooks and virtual machines as unsupported. Wispr Flow covers more languages for transcription (100+ against Qenzo's 42); Qenzo is the one that can output a chosen language different from the one you spoke, per message.

Qenzo vs Wispr Flow at a glance

Axis	Qenzo	Wispr Flow
Where it runs	A system keyboard on Android, so it works in any app without switching tools; an iPhone keyboard is built the same way (requires iOS 17+); hold a key on macOS, Windows or Linux and the text lands where the cursor is	A floating "Flow Bar" overlay on Mac and Windows, opened with a hotkey; a real "Flow Keyboard" on iPhone that has to be switched to via the globe key in every app; a floating "Flow Bubble" on Android, not a keyboard
Platforms	Android 8.0+, iPhone (iOS 17+, no iPad), macOS 11+, Windows 10/11, Linux (X11 native, Wayland via ydotool)	macOS 12+, Windows 10/11, iOS 18.3+, Android 13-16. Their own docs: "iPad, Linux, Chromebooks, and virtual machines or remote desktop environments are not supported"
Price and free tier	Free: 20 actions with no account, 50 more once an email is linked, one-time, no card, no trial. Pro $9.99/month or $79/year on qenzo.ai ($79.99/year in the app stores); most countries pay a lower regional price, down to $1.99/month	Free: word-limited per week, not per account lifetime — 2,000/week on desktop, 1,000/week on iPhone, unlimited on Android. Pro $15/user/month, or $12/user/month billed annually. Growth/Enterprise from $23/user/month (or $18 annual) with SSO and audit logs
Languages and output	42 languages to speak in; 42 output languages chosen per message, so a message spoken in one language can be sent in another. Simplified Chinese, traditional Chinese and Cantonese are three separate choices	100+ languages, including regional variants like British/Canadian English and Swiss German. One language is picked per dictation session, not per message, and Flow transcribes what was said — it does not translate into a different output language
Tone, recipient, role	Seven tones and eight kinds of recipient, chosen by the user at any time; tapping a new tone rewrites the existing text without redictating. 100 professions/roles, up to four active vocabularies	"Flow Styles" auto-picks one of four presets (Formal, Casual, Very Casual, Excited) based on which app you're in, set up in advance and works best in English. "Transforms" (beta, Mac/Windows; Polish also on iOS) rewrites selected text on a shortcut or a custom prompt. No recipient or role concept
Offline behaviour	Typing, word suggestions, autocorrect and Latin-to-native transliteration work with no network; only the AI steps need a connection	No offline mode described. Their docs: "Flow requires an internet connection for voice transcription," and outside iPhone it isn't a keyboard, so there's no typing layer to fall back to
Voice messages in your voice	Pro feature: send the finished text as a voice message in a clone of your own voice, including in a language you don't speak	Not described anywhere on wisprflow.ai
Personal dictionary	Up to 200 names, brands and terms, plus a forbidden-word list and "teach it a phrase"	A dictionary exists ("teach Flow your words"); bulk CSV import up to 1,000 entries per file (3MB max), entries 1-60 characters on Android. No fixed cap stated for entries added by hand
Model choice	Seven model families for the writing step — Gemini, GLM, DeepSeek free; Claude, GPT, Qwen, Kimi on Pro. Speech recognition itself isn't switchable	Proprietary in-house models ("we actually predate OpenAI's Whisper"); no user-facing model choice is described

Where each one actually runs

On Android, Qenzo registers as the phone's keyboard itself, so it opens automatically in WhatsApp, Gmail, Slack or any other text field, with no extra step; an iPhone keyboard is built to work the same way, requiring iOS 17 or newer. The desktop client works the same way in spirit — hold a key, speak, release, text appears where the cursor was.

Wispr Flow differs by platform. On Mac and Windows it's a floating "Flow Bar" opened with a hotkey that inserts text on release. On iPhone it does ship an actual system keyboard, the "Flow Keyboard," but per their own docs you switch to it manually via the globe key, and "iOS resets to the Apple keyboard when you open a new app." On Android there's no keyboard, only the "Flow Bubble," a floating overlay.

Platforms, and specifically Linux

Qenzo's desktop client is released for macOS, Windows and Linux from one codebase: a one-line installer (curl -fsSL https://qenzo.ai/install.sh | sh), plus an AppImage and a .deb, running natively on X11 and on Wayland with the ydotool helper for pasting.

Wispr Flow has no Linux build. Their support docs: "There is no native Linux application. If you run Linux in WSL or a VM on Windows or Mac, use Paste last transcript." Their system requirements page lists Linux alongside iPad, Chromebooks, and virtual machines or remote desktop environments as "not supported." For a Linux daily driver, Wispr Flow isn't an option beyond that workaround; Qenzo runs there directly.

Price and free tier

Qenzo's free allowance is a lifetime count, not a weekly reset: 20 actions with no account, then 50 more (one-time) once an email is linked — 70 total before Pro is needed, no card anywhere in that flow. Pro is $9.99/month or $79/year on qenzo.ai ($79.99/year in the app stores), and most countries outside the US pay less under Qenzo's regional pricing, down to $1.99/month.

Wispr Flow's free tier resets weekly and varies by platform: 2,000 words/week on desktop, 1,000/week on iPhone, unlimited on Android, plus a separate weekly cap on Notetaker. Wispr Flow's Pro plan is $15/user/month, or $12/user/month billed annually. Wispr Flow also sells Growth/Enterprise tiers, from $23/user/month (or $18 annual), with SSO/SAML and audit logs — something Qenzo's site doesn't describe an equivalent to.

Languages, and output language per message

Wispr Flow supports more languages for raw transcription: "100+ languages, including regional variants such as British English, Canadian English, Swiss German, and Simplified or Traditional Chinese." But language is picked once per session, not per message — "Flow picks a language at the start of each dictation session and transcribes the whole segment in that language" — and nothing in their docs describes translating that transcription into a different output language.

Qenzo speaks fewer languages (42) but separates spoken language from output language: speak in one language, choose any of the 42 as the language the message is sent in, chosen fresh each time. Simplified and traditional Chinese are two ways of writing one language and count once; Cantonese counts on its own — Wispr Flow's own docs list Cantonese as "a distinct language option, separate from Chinese – Simplified and Chinese – Traditional" — but only Qenzo's three Chinese options can also be the chosen output language, since Wispr Flow has no translation step at all.

Tone, recipient and role rewriting

Qenzo gives the user seven tones and eight kinds of recipient (boss, client, colleague, friend, family, partner, an official body, or nobody in particular), plus real people the user adds by name, chosen explicitly and changeable after the fact — tapping a different tone rewrites the text you already have rather than re-recording. Author role adds a further axis: 100 professions in one tap, or the user's own description.

Wispr Flow's closest equivalent is "Flow Styles": a per-app-category tone preset (Formal, Casual, Very Casual for personal messages only, or Excited) configured once in advance and applied automatically, rather than picked per message. Their docs note Flow Styles "work best when your dictation language is English." A separate beta, "Transforms," lets you select existing text and rewrite it with a shortcut or a custom prompt on Mac and Windows (a "Polish" preset also on iOS) — useful, but there's no recipient concept and no profession/role system.

Offline behaviour

Neither product's AI dictation works without a connection. Qenzo says plainly it is "not an offline product" for anything the AI does, but its keyboard still functions when disconnected: typing, word suggestions (built-in dictionaries for 25 languages), autocorrect, and Latin-to-native-script transliteration for Armenian, Russian, Ukrainian and Greek keep working with no network.

Wispr Flow's docs state plainly: "Flow requires an internet connection for voice transcription." Because it isn't a full keyboard replacement on desktop or Android, there's no typing layer underneath it to fall back to — without a connection, Flow has nothing to do.

Dictionary, model choice and voice messages

Qenzo's dictionary holds up to 200 names, brands and terms, alongside a list of words Qenzo must never use and a "teach it a phrase" mechanism for corrections that stick. Wispr Flow also has a dictionary ("teach Flow your words"), with a documented bulk-import limit of 1,000 entries per CSV file (3MB max) and a 1-60 character limit per entry on Android; their docs don't state a cap for entries added one at a time.

Qenzo lets the user pick which of seven model families does the writing — Gemini, GLM and DeepSeek free, Claude, GPT, Qwen and Kimi on Pro — while speech recognition itself stays fixed, chosen by Qenzo through measurement rather than by the user. Wispr Flow describes its transcription as running on proprietary in-house models and doesn't offer a comparable model picker.

Qenzo Pro can also turn finished text into a voice message spoken in a clone of the user's own voice, including in a language the user doesn't speak, one tap into WhatsApp. Nothing on wisprflow.ai describes an equivalent voice-cloning or voice-message feature.

Where Wispr Flow is the better pick

If the work is desktop-first — typing into a laptop all day across Slack, Notion, an email client or an IDE — Wispr Flow has more purpose-built tooling for that: documented integrations with Cursor and VS Code, shortcuts that route dictation straight to Slack, email or a calendar entry, snippets, a notes scratchpad, and Transforms for rewriting any selected text on demand, not just what was just dictated. Its 100+ language count beats Qenzo's 42 for anyone who needs a language Qenzo doesn't cover, and for a team that needs seat billing, SSO and audit logs, Wispr Flow has an enterprise tier Qenzo's site doesn't describe an equivalent to.

Where Qenzo is the better pick

If dictation needs to happen on a phone as the actual keyboard — not a bubble tapped into and hoping it inserts correctly — Qenzo is built that way on Android, with an iPhone keyboard (iOS 17+) built the same way. It's also the only one of the two with a native Linux desktop client, and the only one that separates spoken language from output language so the same recording can be sent in a different chosen language each time. Its tone-and-recipient system is picked explicitly per message rather than preset by app category, its free tier has no weekly reset to plan around, and it's the only one of the two that can turn a message into a voice note in the user's own voice.