ArtSeal training-set manifest
Pick the folder you trained on. This page hashes every file on your own computer and gives you a manifest and a Merkle root to put on your listing. The images themselves stay with you — the page has no way to send them anywhere.
Drop your training folder here
or
Nothing hashed yet.
Training-set commitment
- algorithm
- files
- total bytes
- skipped
Give the root and the manifest to ArtSeal at listing time. Keep the folder unchanged — re-running this page over the same files always produces the same root, which is how the commitment stays checkable later.
Files hashed
File names are shown here for your reference only. They are not in the manifest and do not affect the root, so renaming or reordering your files changes nothing.
| name (local only) | bytes | SHA-256 |
|---|
What this page does and does not claim
- It records which exact files you committed to, at a moment in time.
- It does not prove you own them, and ArtSeal never says it does — your attestation is the claim; this is the evidence it attaches to.
- It does not check that these files are the ones your model was trained on.
- The manifest holds sizes and hashes only — no file names, no paths, no timestamps, nothing about your machine.
Verify this page for yourself
The self-test runs the committed spec vectors against the code in this page. Identical vectors are checked by the command-line tool and by the Python implementation, so all three agree or the build fails.