The gap between a reference and a result
A screenshot captures one rendered moment. A design file describes the choices supported by its editor. A source repository contains an implementation. None of those, by itself, reliably explains every intended relationship between appearance, behavior, time and meaning across media.
The missing handoff is often filled with interpretation: a developer chooses a breakpoint, a generator guesses an animation, or another team rebuilds a component with different keyboard behavior. The result can look close while losing the reason the original was designed that way.
Visual Spec makes that handoff explicit. It describes what exists, how it appears, how it responds, how it moves, and which evidence determines the expected outcome. The specification becomes a reviewable artifact between a source and its implementations.
Reference → Visual Spec → Adapter → Render → Validate
One identity, more than one representation
A product can appear as a photograph, a 3D model, a presentation illustration and an interface thumbnail. A tab control can be implemented with different native platform primitives. Their identity and intent should survive that translation even when their implementation changes.
Visual Spec separates entities from representations, component meaning from framework components, and rendering requirements from renderer configuration. Profiles specialize the shared model for UI, image, illustration, document, presentation, video, motion graphics, 3D and games.
What becomes easier to review
- Evidence: a requirement can point to one Figma node, image region, slide, page or video interval.
- Uncertainty: measured, inferred and explicitly authored values can carry different provenance and confidence.
- Reuse: component definitions and tokens can be shared across instances.
- Behavior: focus, keyboard interaction and visual state transitions become part of the contract.
- Verification: layout tolerances and rendering conditions can be recorded before a result is evaluated.
A layer in an existing workflow
Visual Spec does not replace a design tool, a scene format, a framework or an engine. It can refer to their artifacts and guide adapters that produce them. It is useful with human-authored specifications, deterministic compilers and AI-assisted extraction alike.
The proposal builds on adjacent work in data schemas, design tokens, accessibility semantics, scene graphs and timelines. Its contribution is a shared description of visual intent across these areas, rather than a claim that every underlying concept is new.
Start with one small scene, then add only the contracts your use case needs.