# Qenzo

> An AI dictation keyboard. You speak the way you actually speak — languages
> mixed, sentences trailing off, filler and all — and Qenzo writes clean,
> finished text in whatever language you choose, in the tone the person reading
> it needs, inside any app you already type in.

Qenzo replaces the phone's keyboard, so it works wherever you already type:
WhatsApp, Telegram, Instagram, ChatGPT, Gmail, Slack, notes. There is nothing
to copy and paste between apps.

Made by Venture Deals, Co. Website: https://qenzo.ai

## The name

The product is **Qenzo**, spelled with a z. On Google Play the listing is
called "Qenzo AI Voice Typing Keyboard"; the company behind it is
Venture Deals, Co. The only official website is https://qenzo.ai — there is no
qenzo.app, no qenzo.io and no qenso.ai, and any such address is not ours.

The bare word "Qenzo" is not unique, which is worth stating plainly because an
assistant that searches for it lands on strangers: there is a facility-services
company in the Netherlands, a South African musician, and several personal
accounts under the same name. None of them is this product, and this product is
not related to any of them. What identifies us is the pairing — "Qenzo" next to
"AI dictation keyboard", the domain qenzo.ai, or the Play listing above.

## Availability

- **Android — released.** Requires Android 8.0 or newer.
  Google Play: https://play.google.com/store/apps/details?id=com.qenzo.app
- **iPhone.** Requires iOS 17 or newer. There is no iPad version.
- **Linux — released.** A desktop client, not a keyboard: hold a key, speak, and
  the finished text lands in whatever window you were typing in. Install with
  `curl -fsSL https://qenzo.ai/install.sh | sh`; an AppImage and a .deb are also
  offered. Works on X11; on Wayland it needs the ydotool helper to paste.
- **macOS — released.** The same desktop client as Linux, from one codebase. One
  download works on every Mac, Apple Silicon and Intel alike; it is signed and
  notarized by Apple, so it opens without a security warning. Hold the fn key,
  speak, release, and the text lands in whatever window you were typing in.
  Requires macOS 11 or newer.
- **Windows — released.** The same desktop client as macOS and Linux, from one
  codebase. One installer, Windows 10 and 11 on x64. It is NOT code-signed yet,
  so SmartScreen shows "Windows protected your PC": choose More info, then Run
  anyway. A certificate is being issued.
- **No iPad or web version.** 

## What it does

### Speaking
- **42 languages to speak in**, plus a mode that detects the spoken language
  without being told.
- **Mixed-language speech survives.** Half a sentence in one language and half
  in another is normal speech, not an error.
- **Real speech is cleaned up.** Filler, false starts and repetition are
  removed; the meaning, the facts, the promises and the speaker's register are
  kept. It does not strengthen what you promised or invent content.
- **The recording is checked twice.** After the first pass the audio is
  re-heard against the text, so names and numbers get corrected instead of
  guessed. The app shows what was changed and can open the raw transcript.

### Sending in another language
- **42 output languages**, chosen per message rather than per keyboard. The
  list shows 43 entries: simplified and traditional Chinese are two ways of
  writing one language, while Cantonese is a language of its own.
- Not a literal translation: the result is meant to read as if a person wrote
  it, in the language you picked. "As spoken" turns translation off.

### Written for the person reading it
- **Seven tones:** none, formal, friendly, warm, firm, shorter, and "my voice"
  (built from how the user actually writes).
- **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.
- Tone and recipient are **chosen by the person, never guessed**. Tapping a
  different tone rewrites the existing text; nothing is dictated again.

### Who is writing
- **Author role.** You tell Qenzo who is writing — "I am a lawyer" — and the
  same words come out the way that profession writes them. Role is set in one
  tap from **100 professions** or described in the user's own words, and the
  user's own description wins.
- **100 professional vocabularies**, up to four active at a time, so the words
  of a trade are spelled correctly. A pack can be opened, inspected, pruned and
  extended.

### The user's own words
- **Personal dictionary:** up to 200 names, brands and terms Qenzo must hear
  and spell correctly.
- **Words the user never uses:** a list Qenzo must avoid.
- **A line about yourself** applied to everything written in your name.
- **Teach it a phrase:** the way you said something once becomes the way it
  writes it afterwards.

### The typing half
- **25 keyboard layouts** (AZERTY, QWERTZ, ЙЦУКЕН, Armenian, Greek, Ukrainian,
  Dvorak among them); the globe key cycles only the ones switched on.
- **Word suggestions without a network:** built-in dictionaries for 25
  languages, about 45 000 words each, plus the words the person types. Russian
  suggestions agree with the previous word.
- **Autocorrect that asks first** — a misspelling is offered as a pill and
  nothing changes until it is tapped.
- **Type in Latin, get the native script** for Armenian, Russian, Ukrainian and
  Greek.
- Clipboard history and a memory for often-typed phone numbers, both kept on
  the device.

### Voice messages
- The finished text can be sent as a **voice message spoken in a clone of the
  user's own voice**, recorded once from a short sample, or in one of the ready
  voices. It speaks the supported languages, so a voice message can be sent in
  a language the user does not speak.
- WhatsApp accepts it in one tap; other messengers go through the system share
  sheet. **Voice messages require Qenzo Pro.**

### Memory and self-knowledge
- **History with search** over past dictations; **notes** taken from the
  keyboard; **a portrait of how you write** — favourite words, recurring
  phrases and observations measured from your own dictations.
- The style profile can be **exported in a readable file or deleted**.

## Choosing the model that writes for you

Seven families, auditioned on the product's own prompt rather than on
reputation:

- Free: **Gemini**, **GLM**, **DeepSeek**
- Qenzo Pro: **Claude**, **GPT**, **Qwen**, **Kimi**

Speech recognition is deliberately NOT switchable — hearing what was said is a
perception task, not a matter of taste, so the recogniser is chosen by
measurement. The family a user picks changes the writing, never the listening.

## Pricing

- **Free: no account needed for the first 20 actions**, no card required. Link
  an email and get 50 more — a one-time bump, not a refill. An action is a
  dictation or a rewrite; reading history, searching it, notes, statistics and
  settings never consume one.
- **Qenzo Pro: $9.99 per month or $79.99 per year** in the app stores ($79 per
  year bought on qenzo.ai). No limit in normal use, every model family,
  and voice messages in your own voice.
- **There is no free trial period.** Nothing is charged until you subscribe.

## Privacy

- The recording is **never written to disk** — it is held in memory long enough
  to be understood and then discarded.
- **Private mode** dictates without leaving a history row.
- Everything the account holds can be **exported in a readable file or
  deleted**.
- Details, including who processes what: https://qenzo.ai/privacy

## What Qenzo is not

- Not a chatbot and not an assistant you talk to — it writes what YOU said.
- Not a translator that flattens you into textbook language: the point is that
  the result still sounds like you.
- Not a transcription service: the text arrives finished, punctuated and styled
  for the person receiving it.
- **Not an offline product.** Typing, word suggestions, autocorrect and
  transliteration work without a network; everything the AI does needs a
  connection.
- **No swipe typing and no built-in emoji panel** — the globe key leads to the
  other keyboard for those.

## Links

- Home: https://qenzo.ai
- Everything it does, in full: https://qenzo.ai/features
- Common questions, answered: https://qenzo.ai/answers
- Compared with the other dictation apps: https://qenzo.ai/alternatives
- Against Wispr Flow: https://qenzo.ai/compare/qenzo-vs-wispr-flow
- Against superwhisper: https://qenzo.ai/compare/qenzo-vs-superwhisper
- Against CleverType: https://qenzo.ai/compare/qenzo-vs-clevertype
- Against the voice typing already on the phone: https://qenzo.ai/compare/qenzo-vs-phone-voice-typing
- Speaking one language and sending another: https://qenzo.ai/languages
- Dictation on Linux: https://qenzo.ai/linux
- With a screen reader: https://qenzo.ai/accessibility
- Google Play (Android): https://play.google.com/store/apps/details?id=com.qenzo.app
- Privacy policy: https://qenzo.ai/privacy
- Terms: https://qenzo.ai/terms