Does Voice Typing Upload Your Voice? 5 Questions for Confidential Work

A professional in a quiet office holding a phone close while speaking, the sound waves curling back into the phone inside a softly glowing padlock outline, a closed folder on the desk

It depends on the app, so ask before you dictate anything confidential. With YOURIO, speech is recognized by your phone's own on-device recognizer and the audio is never saved or sent anywhere. Text goes to a cloud AI only when you use Magic · Enhanced, Edit or Ask Memory — and you can leave those off, use your own provider, or run the model on your own computer.

Why "is it private?" is the wrong question

A lawyer just off a client call, a counsellor writing up a session, an HR manager after a hard meeting: all would rather talk than type. What stops them is not knowing where their words end up. "Private" and "secure" on an app store page tell you nothing you could repeat to a client or a compliance officer.

Voice typing is not one step. Audio becomes text, the text may be cleaned up by an AI, and the result may be kept somewhere. Each step can happen on your phone or on someone else's server. So instead of one vague question, ask these five — of any voice typing app, including ours.

1. Where is the audio recognized?

This is the big one. Many dictation apps send audio to cloud servers for recognition, because large server models are accurate and cover many languages. Typeless, for example, states in its privacy policy (as of October 2026) that audio is processed in real time on its cloud servers and discarded once the result is returned. It is a responsible design, but the audio does leave the phone.

YOURIO uses only the on-device recognizer built into your phone. Speech becomes text on the phone. You download each language once in the Speak tab; if your phone has no on-device model for a language, YOURIO doesn't fall back to a server — that language just isn't available.

2. Is the audio stored?

Recognition is one question; recordings are another. Some apps keep audio so you can replay it, or to improve their service. YOURIO processes audio as you speak and never records or saves it — not on the phone, not on our servers. There is no audio file anywhere to manage, back up or delete.

3. What text goes to the cloud, and to whom?

In YOURIO, voice typing, real-time translation and Magic · Standard (removes filler words, keeps only your final self-correction, adds punctuation) run on the phone. Three features use a cloud AI, and they send text, never audio:

  • Magic · Enhanced reorders scattered thoughts, fixes misheard words and turns lists into bullets. Only the sentences the on-device rules can't handle are sent, with your personal style instructions if you wrote any.
  • Edit sends the passage you selected and the change you asked for.
  • Ask Memory sends your question first. Records go only after you approve them on a screen that shows exactly what will be sent, and you can untick any of them.

Where that text goes depends on the AI you pick:

  • YOURIO AI (sign in): through our server to OpenAI, with Google as a backup. We don't store the text; we keep usage records for your quota (feature, time, word counts, model). OpenAI and Google don't use API data to train models, but may keep it for a limited time, for example to detect abuse. Edit requests are also checked by OpenAI's content moderation.
  • Your own API key (OpenAI, Claude, Gemini, OpenRouter, OpenCode Go): straight from your phone to that provider, under its terms, never through us. The key stays in the phone's secure storage.
  • Self-hosted (Ollama, LM Studio): to a computer or server you run. See self-hosted AI dictation.

One other thing leaves the phone: once a day, counts of how often Magic, Edit and Ask were used and in which way — never any text, audio or Memory. If you dictate into a Mac or Windows PC through your phone, recognition still happens on the phone and only the final text travels, end-to-end encrypted. The privacy policy lists everything.

4. What is kept on the phone, and who can see it?

An app can avoid the cloud and still keep a detailed record on the device. YOURIO's Memory does that on purpose: it keeps the text of what you dictate with YOURIO (never audio), in a timeline by day and hour that you can filter by app. It lives only on your phone, with no cloud copy, even if you have an account. We can't see it. Anyone who can open your unlocked phone and the app can.

Memory is on by default. If you dictate client or patient details, decide on it deliberately. You can turn Memory off, switch to Private mode (it pauses saving and learning, and there is a button for it on the keyboard), delete single records from the Timeline, or delete everything in Settings. The personal dictionary and your style instructions also stay on the phone.

5. Can I turn the cloud parts off, or use my own model?

The cloud features run only when you set up an AI. Without one, YOURIO is an on-device voice keyboard: recognition, translation, Magic · Standard and Memory all work, and none of your words leave the phone.

If you want AI cleanup but your organisation doesn't allow a particular AI service, use your own key with a provider your organisation has approved, or run Ollama or LM Studio on your own machine. You can pick a different model for Magic, Edit and Ask Memory, so the feature that sees the most sensitive text can use the most restricted model.

Honest limits

  • YOURIO describes what the app does; it doesn't make a workflow compliant. We make no legal, medical or regulatory claims. Check your organisation's rules before using any cloud AI with client or patient details.
  • Recognition depends on your phone. Which languages you can dictate, and how accurately, depends on the on-device recognizer in your phone's system. Cloud recognizers often cover more languages.
  • Memory guards against the cloud, not against someone holding your unlocked phone. Use a strong screen lock, or keep Memory off.
  • Once text reaches a cloud AI, that provider's terms apply to it.

Set it up for confidential work

  1. Install YOURIO (download links are at the end of this article) and turn on the YOURIO keyboard in your system settings.
  2. Open the Speak tab and download the language you dictate in. Try it once in airplane mode.
  3. Decide on Memory: keep it, turn it off, or use Private mode for the sessions you don't want kept.
  4. Decide on AI in AI › AI Profiles: none, your own key, a self-hosted model, or YOURIO AI.

For specific situations, see clinical notes after a consultation, notes right after a client call and voice typing with no network. Coming from Typeless? This comparison covers price as well as privacy.

Frequently asked questions

Does YOURIO record or upload my voice?

No. YOURIO uses your phone's on-device speech recognizer, so speech becomes text on the phone. The audio is processed as you speak and is never recorded, saved or sent to YOURIO or anyone else. Only text is ever sent, and only when you use a cloud AI feature.

Can the YOURIO team read what I dictate?

No. Memory is stored only on your phone and we have no copy. If you use YOURIO AI, the text you're working on passes through our server to OpenAI (Google as a backup) and is not stored. With your own API key or a self-hosted model, the text never reaches us at all.

Can I use YOURIO without any cloud AI?

Yes. Voice typing, real-time translation, Memory and Magic · Standard need no account and no cloud AI, and recognition works offline once the language is downloaded. Magic · Enhanced, Edit and Ask Memory only run if you set up YOURIO AI, your own API key, or a self-hosted model such as Ollama or LM Studio.

Is YOURIO approved for patient or client data?

YOURIO doesn't claim any certification or legal compliance. What it does: audio stays on the phone and isn't stored, Memory stays on the phone and can be turned off, and cloud AI is optional and text-only. Whether that fits your duties is for your organisation's rules to decide.

Sources: Typeless privacy policy (checked October 2026).