Upsurge turns your voice into clean text. This page explains, in plain language, what we collect, why, and the control you have over it.
The short version
We don't sell your data, and we don't use your dictations to train models.
Audio is used only to produce your text, then discarded from our servers.
Your message history lives on your device and in your own iCloud — not on our servers.
We never ask for an email, name, or password. If you subscribe to Pro, Apple handles the payment — we never see your card.
What we collect
Voice recordings — when you dictate, the audio is sent to our service to be transcribed and cleaned. It is processed transiently and is not stored on our servers after your text is returned.
Cleaned text — the finished text is returned to your app and saved to your history on your device (and synced through your own iCloud if you enable sync). We do not keep a copy.
Words Upsurge learns — so it spells the names and terms you actually use the right way, Upsurge notes proper nouns it spots in your dictations (a colleague's name, a product, a place) and keeps them on our servers against your anonymous token. Only the individual words are kept — never the message they came from — and a word is used to help transcription only once it has recurred a few times. You can see the list, remove any word, or clear it entirely in the app.
A device identifier — on first launch the app creates an anonymous token so your requests can be authenticated. No email, name, phone number, or password is required.
Anonymous usage signals — aggregate, non-identifying metrics (e.g. that a dictation happened, its length) so we can improve reliability and accuracy. These are not tied to your identity or the content of your messages.
Subscription status — if you subscribe to Upsurge Pro, Apple bills you and tells us only whether your subscription is active, tied to your anonymous token. We never receive your name, email, or card details; Apple handles all payment information.
How your voice is processed
To turn your speech into clean text, Upsurge uses two AI providers acting on our behalf. ElevenLabs hosts the speech recognition model that transcribes your audio into text. Anthropic provides the language model that cleans up that text and, when you ask for it, translates it — and the same model produces the optional writing feedback and the recap of a period, when you ask for those, from the messages you select. They are bound by contract to process the data only to return your result, to protect it to the same standard we do, and not to use it to train their models. Neither retains your recordings or transcripts. Your recordings and messages are never sold, rented, or shared with advertisers or data brokers.
Storage and retention
Recordings: not retained after processing.
Your history: stored locally on your device, and in your personal iCloud account if sync is on. You can delete any message, or clear everything, from inside the app.
Words Upsurge learns: kept for about four months after your last dictation (about a year for words you add yourself), then deleted automatically. Removing or clearing them in the app deletes them straight away.
Anonymous analytics: retained in aggregate to understand product usage over time.
Your choices
Delete individual messages or your whole history in the app at any time.
Turn iCloud sync off in Settings.
Request deletion of any server-side data tied to your device token by emailing hello@heyvoxly.com.
Permissions
Upsurge asks for microphone access (to record when you choose to) and, on Mac, Accessibility access (to type the cleaned text where your cursor is). It records only while you hold or toggle your dictation shortcut.
Changes
Features and this policy may change as the product evolves. We'll update the date above when we make material changes.