Last updated: 14 August 2026
Short version: we keep your audio for a week, we don’t sell anything to anyone, and the only data that stays longer is the transcription itself — the note positions — which we use to make the model better. Details below, including how to have it deleted.
| Data | Why | How long |
|---|---|---|
| If you sign in: your Google account identifier and email address | to keep your transcriptions in one library and to match a subscription to you | until you delete the account |
| The audio file you upload | to transcribe it and to play it back in the player | 7 days |
| The transcription: notation, MIDI, JSON, tempo, hit count | to show your result; afterwards, to improve the model | kept |
| Corrections you send with the Fixes button, together with that song | to train the model on real mistakes — this is the trade for the tool being free | kept |
| Usage counters: a monthly key derived from your IP address, and the number of songs | to see how much one person uses the service and to enforce fair limits | monthly, then the key becomes meaningless |
| Product events: upload, processing time, whether the player was opened, played, edited | to know which parts of the product are actually used | kept, aggregated |
| If you sign in: your practice calendar and best tempo per rudiment | so the streak and the records follow you to another device instead of starting over | until you delete the account |
We do not store your IP address. For counting we keep a one-way hash of it salted with a secret that changes every month, so after the month ends the value cannot be linked back to an address at all. We don’t ask for your name. An account is optional — the whole service works without one; signing in is what gives you a library that survives changing devices.
On a single server in Nuremberg, Germany. Results are reachable only through the private link you receive — the address contains a random identifier that isn’t published, indexed or guessable.
The landing page and the waiting page can load Google Analytics (where visitors come from) and Microsoft Clarity (how pages are used, including cursor movement and clicks). Both load only after you press Accept in the banner; decline, and no cookies are set and no request is made to them. You can change your mind any time through the Cookies link in the footer.
These scripts are deliberately absent from result pages. The address of a result page is what grants access to it, and we don’t hand that to third parties.
Audio disappears by itself after 7 days. If you have an account, you can delete it and everything in it yourself, immediately, from your account page — see what exactly goes and what stays. Without an account, or if you would rather we did it, write to support@drumanalyzer.com with the link to the result. Under the GDPR you may also ask what is stored about you and object to its use.
One thing deletion cannot reach, and we would rather say it plainly: corrections sent with the Fixes button are merged into the data the model is trained on, and they survive inside the model as numbers. Nothing there is stored per person or identifies you, but it cannot be taken back out either.