Roma reads container credentials first, then samples frames for resolution analysis. The two layers are shown independently.
Learn
C2PA for video
Video credentials live at the container layer. Remuxing and H.264 re-encodes often leave pixels without a Manifest.