This article lists what Sprecho actually holds about you, how long each category is kept, and which servers it sits on. It is written to be quotable in a data protection assessment, so it names specifics rather than making general assurances.
If you want the contractual version instead, the privacy policy and the Trust Center carry the binding text.
What data does Sprecho store?
| Data | Where it is stored | How long, by default |
|---|---|---|
| Audio recording of a dictation | Server in Germany, plus a copy in the app's folder on your own device | 14 days, then deleted automatically. Adjustable from "delete immediately" up to 3 months |
| Transcript: the raw recognition, the formatted result, and any edit you made | Server in Germany | Kept until you delete it. Adjustable, including "do not store at all" |
| Dictation metadata: timestamp, duration, language, writing style, the name of the app you dictated into | Server in Germany | Same period as the transcript it belongs to |
| Account data: email address, display name, interface language, time zone, plan | Server in Germany | Until you delete your account |
| Usage statistics: words per week, dictations per day, time saved, measured typing speed | Server in Germany | Until you delete your account. These are aggregate counters and contain no dictation text |
| Dictionary entries and snippets | Server in Germany | Until you delete them |
| Invoices and bookkeeping records | LexOffice (Germany) | 10 years, required by German tax law (§ 147 AO). These cannot be erased on request |
| Website server log files: IP address, browser, operating system, referrer, time of request | Hosting provider, Germany | 14 days |
Audio and transcripts are two independent settings with two different defaults. That surprises people: recordings disappear on their own after two weeks, while the text stays until you remove it. Both are under your control in Settings → Privacy → Data Retention.
Deleting a dictation inside the app is not a device-only action. It removes the transcript and the audio from the server as well, and the deletion reaches your other signed-in devices through the normal sync.
Where does the data physically live?
All persistent data sits with STRATO AG in Germany (Falkenstein, Nuremberg and Berlin data centres, ISO 27001 certified). Speech recognition and the formatting model run on GPU servers operated by Media Trooper GmbH in Germany and the Netherlands. That step is transient: audio and text pass through for processing and are not retained there.
No customer data is transferred outside the EU/EEA, and no third-party AI provider is involved — inference runs entirely on Sprecho-operated infrastructure. The complete, current list of processors is published in where Sprecho runs and who processes your data.
What happens to a recording while you dictate?
- The app captures audio locally while you hold the dictation key.
- It is sent over TLS to
api.sprecho.ai, hosted in Germany. - The backend forwards it to the speech-recognition model, and the resulting text to the formatting model.
- The formatted text comes back to your device and is inserted into your application.
- The recording is then filed under your audio retention setting — 14 days unless you changed it — and deleted by a nightly job when it expires.
What Sprecho never stores
- Keystrokes outside a dictation, screen contents, your files, browser history or contact lists.
- Analytics or tracking data from the desktop and mobile apps. Neither app contains an analytics SDK. Cookies and measurement apply to the website only.
- Anything at a third-party AI provider. Audio, transcripts and account data are not shared with OpenAI, Anthropic, Google, Microsoft or any comparable service.
Customer audio and transcripts are never used to train Sprecho's models.
The "Help improve Sprecho" switch
Settings → Privacy carries a switch that contributes anonymised text patterns to model quality work. It is on unless you turn it off, and it is worth knowing exactly what it does:
- Names, email addresses, phone numbers and addresses are replaced with fictional stand-ins on your device, before anything is transmitted.
- Only text is submitted. Audio recordings are never part of it.
- Turning the switch off stops submissions immediately.
- In a team, the owner can switch it off for every member at once, and that decision overrides the individual setting.
If you are on a team
A team owner can set a retention policy that applies to every member, and it is enforced on the server rather than in the app. When that is active, the retention controls in your app are read-only and show "Managed by your team". See how Enterprise licences and seats work for what an administrator can and cannot see — notably, an administrator cannot read another member's transcripts.
To change any of this, continue with turning off transcript and audio storage or deleting your data or your account.