Faking Vision

Faking Vision is an operational method derived from Making Vision for representing visual observations in a form that can be reconstructed, edited, tested, and compared.

Rather than treating an image or video as something that must simply be copied or described in an unconstrained prompt, Faking Vision converts visual observations into a structured FV packet containing explicit entities, instances, properties, spatial relationships, and, for video, temporal changes and persistent identities. For images, the core packet consists of an Object Inventory (OI), Map Spatial Relationship Manager (MSRM), and optional Project Constants (PC). Video additionally uses Temporal Data (TD) to make temporal structure explicitly queryable.

The central rule is a closed-world reconstruction principle: the packet is the reconstruction source of truth. A renderer may interpret the representation imperfectly, but it cannot silently invent semantic content that is absent from the packet. Unresolved observations remain unresolved; background context does not automatically authorize expected objects; and persistent identities remain distinct from momentary visual appearances.

This makes the representation itself editable and experimentally manipulable. Objects can be moved, removed, or added; properties and spatial relationships can be changed; and video identities can persist, change state, move, or disappear from view. The resulting representation defines a new reconstruction target rather than modifying an already-rendered image.

Faking Vision also separates source analysis, representation, compilation, rendering, and evaluation. A failed reconstruction can therefore be investigated at the appropriate stage: the visual analysis may be wrong, the FV packet may be wrong, compilation may lose information, or the renderer may fail to reproduce a correct packet. A visually convincing reconstruction is not treated as evidence that its underlying representation was correct.

Faking Vision is therefore a practical companion to Making Vision: Making Vision develops the account of visual processing and selective information allocation; Faking Vision turns relevant principles into an explicit, inspectable representation that can be handed to a reconstruction system and tested against the source.

The method is deliberately concerned with what the representation contains and what the reconstruction system is permitted to infer from it, rather than requiring a particular renderer or rendering technology. The representation is the experiment; rendering is the test.