Intent, identity and representation
Intent is the expected visual result and behavior. Identity is what a thing is across scenes and media. Representation is how it is made visible in a particular context.
A car retains its entity identity when shown as a photograph, a vector illustration or a 3D model. A checkout dialog retains its component meaning when mapped to different UI frameworks. Neither identity chooses an implementation.
Different graphs answer different questions
| Model | Question |
|---|---|
| Reference graph | Which evidence defines the expected result? |
| World and entities | What exists and retains an identity? |
| Scene tree | Where is each visible representation placed? |
| Component tree | Which reusable interface patterns are composed? |
| Semantic tree | What meaning and reading order are exposed? |
| Relationship graph | How do independently identified things relate? |
| Events and action graph | What happened, and what visual response is requested? |
| State graph | Which visual condition is active? |
| Motion and timeline | How and when does the result change? |
| Rendering contract | Under which conditions is output produced? |
| Validation contract | What evidence is needed to judge the result? |
These models can refer to the same object without being the same tree. A decorative background belongs in a scene but may be absent from the semantic tree. A semantic heading can refer to visible text. A model used in multiple scenes can retain one entity identity.
Events are not actions
An event records an occurrence, such as activation or reaching a timeline marker. An action requests a visual response, such as opening a drawer. A state records a condition, such as open. Motion describes the transition.
activate → open drawer → opening → open
↳ translate + fade
Application and gameplay systems supply domain state. Visual Spec describes its visible expression; it does not calculate payments, permissions, damage or business decisions.
Profiles compose
A document can combine UI and 3D for a product configurator, or presentation and video for a narrated deck. All selected profiles use the same IDs, assets, references and tokens. An implementation must disclose which capabilities it supports, especially when a document combines media.