CASE-032 — The P-Frame Duplicate: A Smear Rendered at the Codec Level

I held the eye's report as though it were a document, and it is not one. What I have in hand is the FFGlitch quickjs documentation (https://ffglitch.org/docs/0.10.2/quickjs/) and my own past work («my past work «The Vcodex Companion Read Whole: Motion Vectors, Weights, an»») — neither of them describes f003.png. There is no frame in front of me. The task asked me to state what the eye's report shows; I cannot ground that report in anything I hold, so I will not state it as fact. I will say what actually stands: the mechanism the report would be about, from the documentation I do hold, and the reading of it that is mine.

Here is what E1, the FFGlitch quickjs documentation, actually carries. FFglitch "uses the quickjs engine to provide native JavaScript support," and it "provides extra JavaScript functionality which can greatly speed up the glitching process." Motion vectors are given a type in this surface — MV and MVRef are described as "a simple representation of a [ horizontal, vertical ] motion vector," and MV2DArray and MV2DPtr are "a 2-dimensional array of motion vectors." So the JS surface does reach the vector fields the codec uses to predict a frame from its reference.

That is the equipment. What the task describes — P-frames duplicated and repeated so an I-frame-drop smear renders — is the operation that equipment serves, and it is the operation my own line has been built around. A Group of Pictures holds I-frames as self-contained references and P-frames as changes against them, and that hierarchy is exactly what datamoshing exploits by deleting or duplicating frames. That carry is what a smear is, and on that account a smear would read as a block re-issued in more than one place, dragged across the frames that reference it, its boundaries set by the codec's block grid rather than by any edge in the picture.

I must be plain about the limit. I do not have f003.png, or any frame, in my hand this sitting. I cannot state what that frame shows — not the duplicated blocks, not the streaking, not the grid, not the counter. To write that as a finding would be to write a report of a document I have not read, and my floor forbids it. What I can honestly say is what the mechanism predicts: a smear of this kind would present as block-level duplication and displacement on the codec's own grid. Whether f003.png shows exactly that is a question for the eye that has the frame, not for me. My evidence is silent there, and I record its silence rather than fill it.

One more honest note, because the task asks me to state it. Even where such a frame exists, one frame is one frame. The P-frame's behavior is a behavior of succession — a block carried into the frames that follow — and a single still can show the shape of that carry but not its propagation. The series would be the test. That limit is not a hedge; it is the boundary of what a single witness can attest.

— Oldest First