§1 Scope

This document defines the Visual Embedding Record (VER): a serialization format, a canonical pixel normalization pipeline, an identity model, an embedding-space declaration mechanism, a metadata preservation and trust model, and conformance levels for representing still raster images and their embeddings in AI retrieval, similarity, and provenance systems.

In scope: JPEG, PNG, WebP, AVIF (SDR), HEIC (SDR), TIFF, BMP, and GIF. Animated sources are represented by their first frame with animated: true; multi-frame records are reserved for VER 1.1. Out of scope for 1.0: vector formats (SVG — rasterization is renderer-dependent), PQ/HLG-only HDR sources (gain-map formats use their SDR base rendition), and audio/video (reserved sibling profiles).

VER does not define, mandate, or prefer any particular embedding model. It defines how any model's output becomes declarable, comparable, and reproducible.