Data Retention & Deletion

Each stored object carries an explicit expiry date recorded alongside it.

• Incomplete uploads — 24 hours. A reservation whose transfer never finished is cleared along with any bytes that did arrive.
• Anonymous sessions — 24 hours from upload, unless you sign in first. Signing in claims the session's uploads onto your account automatically and re-stamps them with your plan's window, as described below. Left unclaimed they are deleted: the guest identity lives only in this browser, so nothing else can ever reach those files again.
• Free accounts — 7 days from upload, for both source assets and everything generated from them. A result is given the same expiry as its source, so neither outlives the other. Past that date the API stops issuing download links and refuses to start new jobs against the asset.
• Starter — 30 days from upload.
• Pro — 90 days from upload.

Retention is stored as a date on the record rather than baked into the file's storage location, which is what lets it change when a session becomes an account. Sign in before an anonymous upload expires and it is claimed onto your account, re-stamped with your plan's window — measured from when the file was uploaded, not from when you signed in.

The window is fixed when a file is uploaded, from the plan you were on at that moment. Ending a paid subscription does not shorten the window on files you uploaded while it was active — those keep the date they were given. Anything you upload afterwards gets the free-account window.

After an upload is validated, we render a small preview image of the model so the asset library has something to show. It is a shaded image of the geometry — 512 × 512, no textures — stored in the same private location as the model itself and never made public. It is fetched through a short-lived link that is checked against your account first, exactly like the model.

The image belongs to the asset: it is deleted when the asset is deleted, expires when the asset expires, and does not count toward your stored-asset allowance. Rendering one costs no processing credits.

The 3D viewer and the 3D GIF maker open your model in the same workspace everything else uses, so the file is stored and the windows above apply to it exactly as they do to any other asset — including the short anonymous one, if you have not signed in. What those two tools never send is the work: the model is drawn, measured and encoded by your own browser, no server renders anything, and neither costs a processing credit. The file converter is the one that does run on our servers.

Assets nearing expiry show their date in the asset library and in the workspace, in the form Expires 30 Aug 2026. Nothing expires silently without that date being visible first.

A scheduled sweep runs hourly and deletes what has expired. For each expired asset it removes:

• the uploaded source file,
• every optimized mesh, LOD level, archive, and companion file generated from it,
• the preview image rendered for the asset library,
• and any artifact a job wrote but failed to record, found by listing the asset's own storage location one final time.

A record is marked deleted only after its files are actually gone. If storage refuses a deletion, the record keeps its identity, its expiry is pulled back so the next sweep retries it, and the failure is logged. Anything still uncleaned a day past its deadline is reported as an operational fault rather than retried in silence.

Deleting an asset from the library is a real backend operation, not a hidden row, and it happens as part of the request rather than on the next sweep. It removes the same set of files listed above, plus the records of the processing jobs that produced them.

A deleted asset stops being reachable immediately, and no new job can be created against it. If any file cannot be removed at that moment, the asset is not reported as deleted — you get an error, and it is marked for another attempt, so the record never claims something is gone while it is not.

If you buy a paid plan, we keep the link between your account and your Paddle subscription: the customer and subscription identifiers, the price, the status, the billing period dates, and the billing email. These are kept after the subscription ends, cancelled ones included — they are the record of what you were charged and what access you had, and deleting them would leave a billing dispute with nothing to check it against.

They contain no card or bank details; those never reach us. Paddle holds the payment records themselves under its own retention policy. See Subprocessors.

Anonymous sessions are free to start, so a limit attached to one limits a browser rather than a person. To stop the same visitor taking an unlimited number of them, we store a one-way scrambled fingerprint of the network each anonymous session was started from — and of the network a contact form was submitted from.

The IP address itself is never stored, and the fingerprint cannot be turned back into one. It is compared only against other fingerprints, to apply the limits published on this site. It is never used to identify you, to build a profile, or to recognise you across visits.

It is kept for as long as the session record it sits on, which is indefinitely — that record outlives the files it owned, because its identifier is part of where those files were stored. Contact-form fingerprints are kept with the enquiry.

Deleting every asset removes all uploaded content; removing the account record itself is handled by support on request. To have the account record removed, email support@meshcrunch.com; we action such requests within 30 days.

Uploaded models and generated outputs are not backed up. Object storage holds exactly one copy, and a deletion — whether yours or the automatic sweep's — is immediate and unrecoverable. Keep your own copy of anything you care about.

The database holding account, asset, job, and subscription records is backed up: a nightly encrypted-in-transit dump to separate private storage, kept for 30 days. That protects your account and the record of what you uploaded. It does not protect the model files themselves — the database holds metadata only, because model data never passes through our API.

So a database failure would cost you at most a day of recent activity. A storage failure, or a deletion, would cost you the files.

The practical consequence for you is simple: keep your own copies. MeshCrunch is a working area, not an archive.

See Privacy for what data is collected and Security for how it is protected.