Privacy, in plain words.
A keyboard and a dictation app see a lot. Here is what ours do with it, app by app. This page is a summary; each app's full policy is linked at the end and always takes precedence.
The short version
Keyboard typing is processed on your phone. Your voice goes to the speech engine set in the app, for the dictations and recordings you start. Nothing is sold, and there are no ads.
Keystrokes stay on your phoneThe iPhone and Android keyboards don't send what you type to us. The two narrow exceptions are below.
Cloud AI: phones ask, desktops are onThe phones ask before using cloud AI. On Mac and Windows, cloud speech and clean-up are on from the start, and each is a switch.
No ads, no tracking, never soldYour data is not used for advertising.
What you type
On iPhone and Android, typing, swipe, autocorrect, emoji and (on Android) Telugu suggestions run on your phone, and the keyboards work offline. They don't send your keystrokes to us. Two things do leave the phone: typing corrections you teach the keyboard are saved to your account when you are signed in, and, with AI features allowed, words you type into the emoji search box may go to Google Gemini to find emoji (on Android, GIF and sticker searches also go to KLIPY).
On Mac and Windows, VoiceInput listens for its shortcuts but doesn't save or send your keystrokes. Screen context, below, can send the text near your cursor along with a dictation.
Your voice
- On iPhone and Android, only after you allow it. On iPhone, your voice goes to VoiceType's speech service only after you allow AI features. On Android, VoiceType asks before anything is sent to an AI service and shows exactly what goes where. Turn AI features off and the phone's own recognizer does the work.
- On Mac and Windows, cloud speech is the default. While you dictate, VoiceInput streams your microphone audio to the speech engine picked in Settings (ElevenLabs Scribe v2 Realtime by default; Microsoft, Meta, Google Gemini, OpenAI or xAI Grok if you pick them), which sends back the text. There is no separate AI consent screen on the desktop. Pick Built-in dictation in Settings to use the computer's own recognizer; it also takes over when the network is down. On a Mac with macOS 14 or 15, Apple's recognizer may still use Apple's servers until macOS has its on-device speech model.
- Only while you dictate, with one Windows exception. The microphone is used while dictation or a recording you started is on. On Windows, after a dictation the microphone stays open for up to 45 seconds so the next one starts instantly. Only the last second or so is held, in memory, and it becomes the start of your next dictation if you begin one in that time; otherwise it is never saved or sent. Turn off Keep microphone ready in VoiceInput's Settings to close the microphone right away.
- Dictation keeps text, not audio. Every app keeps the words you said, not the audio of a dictation. Recordings you start with VoiceInput's capture shortcuts are different; see below.
Clean-up
Each app has separate clean-up switches: custom vocabulary, fixing obvious grammar, a tone and, on Mac and Windows, a User prompt. When a step runs in the cloud, the text of that dictation (with your vocabulary and instructions) is sent to the AI model that tidies it; the audio is not. The result is only ever a tidied version of your own words.
- Android: every clean-up step is off until you turn it on.
- iPhone: cloud clean-up runs only after you allow AI features.
- Mac and Windows: Grammar and Custom vocabulary are on by default (Tone and User prompt are off), so unless you turn them off the text of each dictation goes to the clean-up model picked in Settings, Qwen on Cerebras by default.
Screen context on Mac and Windows
Screen context is on by default. Whenever a clean-up step runs, VoiceInput reads the app's name, the window or tab title, the text around your cursor in the field you are typing in and the text visible in the window (read from the screen), and sends it with that dictation to the clean-up model, so names, code and terms match what you are working on. It never reads a password field. VoiceInput does not keep it; the AI provider processes it under its own terms. Turn it off in Settings › Post-processing. The Java code shortcut (⌥J / Alt+J) always reads the screen.
Recordings, meeting notes and screenshots on Mac and Windows
Recordings, meeting notes, screenshots and adding tasks need you to be signed in (on Windows, reading selected text aloud also works signed out). VoiceInput saves their results as tasks in your account's to-do list, the same list the To-Do List app shows.
- Recordings and meeting notes capture your microphone and your computer's sound, which includes other people on a call. Tell the people you record, and get their consent where the law requires it.
- Recordings are transcribed by your account's recording engine (ElevenLabs by default, set in the To-Do List app), which can differ from the speech engine you dictate with; a single-speaker recording also streams live words to Microsoft unless ElevenLabs is your speech engine. Titles and notes are written by your Notes AI model (Qwen on Cerebras by default, with Google Gemini as its fallback). The original recording is kept in your account unless you turn off Keep the original audio (an audio-only recording is always kept). Meeting notes are streamed live to Meta's speech engine and kept as text.
- Screenshots are read by Google Gemini, and the image is saved with the task.
- Read aloud sends the selected text to Microsoft Azure text-to-speech. Add as a task saves the selected text to your list.
Your account
- On the phones an account is optional. On Windows, dictation works without one; the capture shortcuts need one. On the Mac, signing in with Google or email is how your VoiceType subscription unlocks the app.
- With an account, your email, custom vocabulary, settings, User prompt, the text of your dictation history and the tasks, recordings and screenshots from VoiceInput's capture shortcuts are stored with Google Firebase so they follow you between devices.
- On iPhone and Android, you can delete your account and its VoiceType data from inside the app at any time. Recordings, screenshots and tasks made with VoiceInput are not removed that way: they are deleted when you email sharathck3@gmail.com from your account's address. VoiceInput for Mac and Windows has no delete button: delete the account in the VoiceType phone app, or email us and we will delete the account and all its data. Android users can also request deletion on the web: delete your Android account.
Purchases
Subscriptions are handled by the App Store (and, on Android, Google Play). We never see your card details; the apps only check whether your subscription is active.
This website
This site uses no cookies, no analytics and no trackers. Fonts are loaded from Google Fonts, and a small release-info script and the desktop downloads are served from Google Cloud Storage.
The full policies
Terms: Terms of Use (iPhone) · Android Terms.
Questions about privacy? Email sharathck3@gmail.com.