Does Voice Typing Send Your Audio to the Cloud?

Two mic buttons can run two different pipelines

Nothing on screen tells you where audio goes. The same microphone icon can mean a model running on your device or a stream heading to a data center, and both feel identical while you speak.

• On device: a downloaded model turns audio into text locally and nothing is uploaded for that step.
• Cloud: audio, or features extracted from it, goes to a server that returns text.
• Mixed: local transcription with an optional cloud request for cleanup, or the reverse.

Ask about the engine selected in your settings right now, and about what the cleanup step does once the transcript exists. Those two answers describe your actual exposure; a general claim about the app does not.

The matrix

Here is where the common paths land. Check the row that matches the tool and setting you are on today, because several of these products offer more than one path.

Note the first two rows carry the same brand and behave differently. Auri Voice and Auri Keyboard are separate apps on separate platforms, and the defaults reflect what each device can do.

Apple and Google, with the nuance intact

Both platforms have shipped on-device speech recognition, and both still use servers in some situations. Summaries that flatten this into “Apple is local, Google is cloud” will mislead you at the moment it matters.

• Language packs and device generation change whether the local path is available.
• Some features layered on dictation, such as server-side assistance, use the network even when transcription does not.
• The switch happens quietly, with no interface change to warn you.

Read the platform's own dictation privacy page for the OS version you are running. If you need a guarantee rather than a general tendency, use a tool where you pick and download the engine yourself.

Cloud AI dictation apps

The apps that clean up rambling speech into finished writing usually run large models on servers. That is how they get their quality and their speed on modest hardware, and it is a reasonable trade as long as you make it knowingly.

• Expect microphone permission plus an account or an API key.
• Expect two requests per message: one for audio, one for the cleanup on the text.
• Retention varies widely. Some discard audio immediately; others keep samples for quality work.
• Airplane mode is the quick test. Cloud-only tools stop working with no network.

Before you dictate client material into one of these, read the audio section of the policy rather than the homepage.

Transcription and cleanup are separate uploads

This is where most confusion lives. People assume a local transcript means nothing left the device, and for many tools the polish step sends the text straight out again.

1. Local transcription, cloud cleanup: audio stays, text leaves.
2. Cloud transcription, local editing: audio leaves, nothing more is sent.
3. Both cloud: one combined pipeline, common in AI dictation apps.
4. Both local: nothing leaves for either step. Auri Voice on Mac works this way by default, removing filler, false starts, and punctuation problems with local processing.

When you ask a vendor about privacy, ask about both steps by name. A single yes or no about “the cloud” cannot describe these four cases.

Prove it with airplane mode

Documentation tells you the intent. A five-minute test tells you the behavior of the settings you actually have.

1. Download any local model the app offers and select it explicitly.
2. Turn on airplane mode and switch Wi-Fi off.
3. Dictate a paragraph and see whether text appears.
4. Try the rewrite or cleanup feature separately and note whether it fails.
5. Reconnect and check the settings screen for anything that switched back on its own.

Write down what worked offline. That single line, kept somewhere you will find it, answers the question faster than rereading a policy in six months.

Rules for the drafts that matter

Not every message needs the same treatment. Sort by content and the decisions get easy.

• Health details, legal matters, unreleased plans, and client names: local transcription and local cleanup, which on Mac means Auri Voice with a downloaded engine.
• Ordinary work email: cloud is fine if you know the retention terms.
• Anything you cannot verify: type the sensitive part by hand and dictate the rest.
• Passwords and card numbers: type them, and only on the system keyboard.

If you never look this up again, keep one habit: check which engine is selected before you dictate something you would not email to the vendor.

FAQ