Audio
Decoded, analyzed and converted on your machine with bundled FFmpeg libraries. No sample ever leaves it — not for analysis, not for 'cloud processing', not for training.
/// Trust pages · privacy & offline
Analysis and conversion run entirely on your machine. No audio and no metadata about your music ever leaves it. This page spells that out — and, because honesty beats a slogan, it also discloses the one narrow feedback channel planned for the beta instead of pretending it doesn't exist.
§ Offline by construction
The whole engine — decoding, every check, every conversion — is code running on your computer against your files. It needs no account, no login, and no network: unplug the machine and every feature keeps working, forever. That is a design constraint, not a temporary state.
Decoded, analyzed and converted on your machine with bundled FFmpeg libraries. No sample ever leaves it — not for analysis, not for 'cloud processing', not for training.
Filenames, folder paths, artist/title/album tags, artwork and playlist names stay local. They appear in your reports and nowhere else.
Check results and the analysis cache live in plain local files (cache under ~/.cache/digwash, settings under ~/.config/digwash). You can read them, and deleting them deletes them.
DigWash reads the files, folders and playlists you explicitly hand it. There is no background indexer, no folder watcher, no scan of the rest of your disk, and no hosted copy of your library.
§ Telemetry, stated plainly
Today's beta builds send nothing on their own — the only network touchpoints are survey links that open in your browser when you click them. But one feedback channel is coming during the beta, and we would rather describe it precisely before it ships than surprise you after.
When DigWash makes a claim — "quality low", "fake lossless suspected", "won't play on X" — you can confirm or dispute it in the tool. That label, plus minimal usage counts, is the field data that tunes the checks: "the tool said low, the DJ said it sounds fine" is exactly the correction an algorithm cannot generate for itself. A clean report asks for nothing.
Events carry a random install ID generated locally — not derived from your hardware, resettable by you at any time, and never stored anywhere that also holds an email address. It exists only so that ten disagreements can be told apart from one very persistent one.
Every event is mirrored to a local file before it is sent, and the tool will show it to you on request. What we see is exactly what you can see — no hidden second payload.
During the beta, accepting this feedback loop is a condition of beta participation, stated in the Beta EULA and in a first-run notice — nothing transmits before that notice has been shown. That is a fair trade to name out loud: beta builds exist to be corrected. At V1 it becomes a real choice: an explicit opt-in/out screen on first run and a settings toggle, with no nagging. Sending never blocks or delays a wash, and the product stays fully functional with zero network either way.
§ Leaving
No account to close, no server-side library to export, no activation to revoke. Delete the app (or the CLI binary) and you are done. Washed copies and playlists are plain files that were always yours; if you want every trace gone, also delete the two local folders — ~/.cache/digwash and ~/.config/digwash — and DigWash has forgotten you completely.
This page is about the app. The website's own (cookie-free, aggregate) analytics are described separately on the website privacy page.
§ Continue