Verify

Verify any asset's provenance.

Wakemark records provenance the moment a generated asset is persisted: a byte-exact sha256, the model and provider that produced it, any embedded C2PA Content Credential, and the lineage it derived from. This page reads that record and shows it in plain language — the human face of GET /v1/assets/{id}/provenance. No account, and no green checkmark unless there's something real behind it.

Provenance JSON

Nothing is uploaded — the document is parsed entirely in your browser. Your prompt/input is never displayed.

Verify an asset's provenance: its C2PA manifest status, byte-exact digest, the model and provider that made it, and the lineage it derived from — with no green checkmark unless there's something real behind it. New here? to see it work.

How verification works

Wakemark's stance is preserve → detect → log: it never transforms bytes, so whatever a provider embedded survives, and it records a digest and an append-only ledger alongside. Three independent signals show up here, each stated for exactly what it is.

C2PA manifest: A signed metadata manifest (Content Credentials) that some providers embed. The record carries four honest states: verified, present but not verified, none, and not inspected. And when you drop the original file, its manifest is validated in your browser— the signature checked and the signer named — so even a record that says “not inspected” can come back verified (or invalid). A missing manifest is never dressed up as a pass.
Byte-exact digest (sha256): The content hash recorded at generation time. Drop the original file and your browser re-hashes it locally (nothing is uploaded) and compares — a match proves the file is byte-for-byte the recorded asset; a mismatch means it was altered or is a different file. This is a real check the page performs, not a claim from the JSON.
Lineage: The derivation chain — the parent assets an output was built from (image → video → dubbed video). Parents are recorded as asset ids; each parent's own provenance is verified the same way.

Three levels of proof

Unverified document
Paste a provenance JSON or sidecar and it's parsed and shown — but a document you supply is only a claim. It proves nothing on its own.
From Wakemark’s records
Look an asset up by its id and the provenance is read from Wakemark's own API — the authoritative record, not something you handed in.
Byte-confirmed
Drop the original file: your browser re-hashes it and compares to the recorded digest. A match is the real proof — math on the actual bytes. This is the point of the page.

Verifying a digest or a Content Credential is a technical check on the bytes and their embedded metadata. Wakemark is designed to support EU AI Act Article 50 marking and logging workflows; nothing here is legal advice, and this page does not by itself make any system compliant with any regulation.

Don't have a record to check? Browse the models & prices or the providers we route across.

Invisible watermarks: what we can — and can't — check

Some generators embed an invisible, statistical watermarkin the pixels or audio themselves — Google's SynthID, Meta's Stable Signature, and others. Unlike a C2PA manifest or a sha256, these are not publicly detectable: only the issuer holds the detector, and the mark is designed to survive edits that would break a hash.

So Wakemark does not fake a check it cannot perform. This page reports what it can actually prove — the byte digest, the C2PA signature, the recorded provenance — and says plainly when a signal is out of its reach. If your asset may carry an invisible watermark, verify it with the issuer's own tool.