Measured 2026-09-11 on the Linux dev server, ffmpeg 6.1.1-3ubuntu5. Reproduce
with bash measure.sh <outdir>; raw output in results.json.
What was measured#
| Question | Result |
|---|---|
Does a CUSTOM metadata key (c2pa_manifest=...) get written into MP4 at all? |
No |
Does a RECOGNISED key (comment) survive a re-encode + scale to 160x120? |
Yes |
| Does it survive a stream-copy remux? | Yes |
Does -map_metadata -1 remove it? |
Yes (one flag) |
| How much payload survives a transcode? | 8000 characters, intact |
| Is a stream-copy remux byte-identical to its source? | Yes |
| Is a transcode byte-identical to its source? | No |
| After MP4 -> WebM, is the key still readable? | Yes, but the case changes (comment -> COMMENT; ffprobe resolves either) |
What follows#
A manifest that survives a transcode reports TAMPERING. The manifest binds
the SHA-256 of exact bytes, and a transcode produces different bytes by
definition, so a surviving manifest verifies as media hash mismatch. That is
the same result a verifier gives for a maliciously edited file. Carrying
provenance through a transcode therefore does not preserve provenance; it
manufactures a tamper signal.
So "survives" is the wrong goal. A remux is byte-identical, so the manifest stays valid and must be carried. A transcode needs its own manifest, issued over the new bytes, whose edit assertion names the parent's media hash. An inherited manifest must be replaced, never passed through.
Two secondary findings shape the transport:
- A custom tag key is silently dropped by the MP4 muxer, so a manifest cannot
ride in one. The recognised
commentfield does carry 8000 characters through a full re-encode, which is enough for a manifest but is a field anything else may also write. -map_metadata -1removes it, and that flag is in most sanitising pipelines. Embedded metadata is therefore a convenience, not a guarantee; the sidecar is the authority.