Questions
The four moments dictation touches the network
"Works offline" is usually true and rarely complete. Here is every point at which a local dictation app uses a connection, and whether it matters.
Last updated
The four moments
| When | What it carries | Required? | Noticeable offline? |
|---|---|---|---|
| First-run model download | Nothing from you — the app fetches a model file | Yes, once | Yes, if you skipped it |
| Licence activation | A licence key and a device identifier | Only if you bought a plan | Once, at activation |
| Update checks | A version number | No | No — fails quietly, retries later |
| Optional cloud rewrite | The text of your dictation | No, off by default | Yes, if you turned it on |
| Product analytics | Usage counts, no content | No — can be turned off | No |
| Recognition itself | Nothing | Never | No |
The row that matters for privacy is the fourth. The row that matters for travel is the first.
1. The model download
A few hundred megabytes to a couple of gigabytes, once, on first run. In Vaitly Voice it is verified by checksum before it is ever loaded.
Do it on a connection you trust, at a desk, before you need it. An app installed but never launched has no model and will do nothing offline — which is the single most common way this goes wrong.
2. Licence activation
If you buy a plan, the licence is validated once with a connection and then works offline. Vaitly Voice Free needs no sign-in at all, so there is nothing to activate — worth knowing if you are travelling imminently.
3. Update checks
Periodic, carries a version number, and fails quietly when there is no connection. Vaitly Voice has no automatic in-app updater: updates are downloaded from the website, which is a genuine rough edge rather than a design choice.
4. The optional cloud rewrite
This is the one to understand properly. Many dictation apps, including ours, can tidy a dictation into cleaner text. If that step is pointed at a cloud provider, the text of your dictation is sent to that provider on every dictation that uses it.
What about cloud dictation apps?
Entirely different answer: a connection is required for every dictation, because the recognition happens on their servers. Wispr Flow's privacy page states that transcription always happens in the cloud; Google Docs voice typing and Microsoft's Dictate both require a connection.
That is not a criticism — it is the architecture, and it buys them accuracy and platform coverage. It does mean they stop working on a train, and stop mid-sentence rather than gracefully.
Does dictation use mobile data?
With an on-device app, effectively none after the first run — a licence check and an update check are a few kilobytes. With a cloud app, every dictation uploads audio, which over a working day is not trivial on a metered connection or while roaming.
If you tether regularly, this is a real and rarely mentioned difference.
Questions
Can I dictate with no internet at all?
Yes, with an on-device app that has already downloaded its model. Recognition needs no connection. Install and dictate one sentence before you travel — that is the only step that requires one.
Does VV need internet to work?
Once, for the model download. After that, dictation, vocabulary, snippets and the tidy-up all work offline. Free dictation needs no sign-in, so there is no licence check either.
Does dictation use a lot of data?
An on-device app uses almost none after the first-run download. A cloud app uploads audio for every dictation, which adds up on a metered or roaming connection.
Will the AI tidy-up work offline?
In VV, yes — the writing model is local. In many apps that stage is a cloud API call and will fail offline. Test it specifically rather than assuming.