Observer
Observer Nuance models inter-observer variation without changing the canonical identity or canonical Lab.
Vision
Give vision a stable color vocabulary.
A camera produces device values. La Nuance adds a matching layer that can associate observed color with stable Registry identities.
Machine vision
World→Camera→Image→Match→NUANCE IDs
Matching is a defined technical step. Capture, calibration, illuminant and imaging conditions still matter; raw RGB is not treated as canonical identity.
Perception layers
Metamerism
Viewing-condition effects can be described separately from the Registry identity.
AI Vision
Structured IDs provide a bounded vocabulary software can index, compare and transmit.
Spectral layer
Modelled reference. Not a physical master.
R4 contains modelled spectral representatives. They are computational references and are non-authoritative for canonical color identity. Physical claims require producer measurement.
MODELLED_SPECTRAL_REPRESENTATIVE__NON_AUTHORITATIVE_FOR_COLOR_IDENTITY
Scope
A vocabulary, not a compression claim.
Compact Token v1 is a 12-bit mapping for the current Registry. The Protocol does not claim that this mapping alone is a complete image codec or automatically reduces model tokens, bandwidth or energy.