The Scene Document

The Scene Document is CineCode's intermediate representation (IR). It's what cinecode-script produces after loading a project, and what cinecode-render consumes to draw frames. Everything in between operates on this one data model.

What it is

A Scene Document is a fully serializable description of a film: scenes, the nodes within them, their animation tracks, the camera, transitions, and references to assets. Because it's plain data (JSON-serializable), it carries no hidden runtime state — which is exactly what makes rendering deterministic.

Why have an IR at all

Putting a data model between authoring and rendering decouples them:

• The front end (Rhai .ccs scripts, the TOML manifest) only needs to produce a Scene Document. New authoring surfaces could be added without touching the renderer.
• The back end (CPU and GPU renderers) only needs to consume one. Either backend draws the same document identically.
• Tools in between — validators, optimizers, analyzers — can read and transform the document without re-running scripts.

Nodes and tracks

Within a scene, every visual element — text, code, a shape, an image, a running simulation — is a node. Each node carries animation tracks: timed keyframes (with easing) describing how its properties change over the scene.

This uniformity is deliberate: a line of code and a terrain simulation are the same kind of thing on the timeline, which is why both can be animated, focused, and linked together. See Simulations as nodes.

Evaluation

At render time the document isn't replayed step by step. Instead, for a given frame n, the timeline evaluates each node's tracks to a concrete value at n. The result is an AnimatedState (plus a CameraState) — a snapshot of exactly what the frame should contain, ready to draw.

That evaluate-then-draw split is covered in The rendering pipeline.