File

Make creative intent legible — to people, pipelines, and large language models.

Customer Story Studio is built around MovieLabs OMC-JSON v2.8: a structured way to describe narrative + production context, assets, versions, and provenance. We’re building a lightweight place to learn OMC, validate real files, and explore how an ontology becomes usable by an LLM.

OMC-JSON v2.8-first Schema + best-practice linting Graph + concept packages Built in a Gemini hackathon Human + LLM readable

Open the viewer Learn OMC 2.8

Why OMC 2.8

“If it loads, it’s reviewable.”

The core idea is simple: creative work becomes collaborative when it’s structured. OMC (Ontology for Media Creation) gives a shared vocabulary for narrative objects, production objects, assets, versions, and provenance — so tools can interoperate and teams can hand off work without re-inventing formats.

What “OMC-strict” means here

Schema-valid OMC-JSON v2.8 data (the admission ticket)
Resolvable references (IDs link cleanly through the graph)
Best practices that reduce ambiguity for tools & reviewers
Extensions via customData without breaking compatibility

Why an ontology viewer matters for LLMs

LLMs produce “content”; OMC captures context.
Structure turns generations into inspectable packages (scenes, characters, props, assets).
A visible graph lets humans spot gaps; a typed ontology lets an LLM reason about what’s missing and what should connect.
Validation gives you a hard edge: invalid files don’t ship.

Built for two core personas

One shared constraint: valid OMC.

1) Emergent AI Generators

You can create fast — but your work becomes “real” to collaborators when it’s packaged as structured intent, not chat logs and folders.

Turn generations into industry-legible objects (scenes, characters, shots, assets)
Keep provenance (what came from where, and why)
Share as a file that opens anywhere (no platform lock)

2) Existing Industry

You need signal, safety, and interoperability — not unstructured AI sludge. OMC is a gating mechanism: structure becomes the filter.

Evaluate work via typed context, not hype
Reduce risk with lineage + versioning
Bring new talent in without contaminating pipeline formats

The Gemini-hackathon build

Viewer + learning playground

Validate

Make correctness non-negotiable. Validation is the first step toward interoperability and trustworthy automation.

Render

Convert JSON into a navigable graph: narrative nodes, production nodes, assets, and their relationships — so a human can actually review it.

Teach

Help an LLM (and the user) understand the ontology by showing “what belongs where,” and by surfacing missing links and incomplete packages.

Try the idea locally:

Status: This static page doesn’t validate against the full OMC schema yet. The real app should wire this to an OMC-JSON v2.8 validator + best-practice checks, then render the ontology graph and concept packages.

Tip: The “secret” is the constraint: only valid OMC gets to participate. That’s what makes collaboration low-friction later.

Quantifying “reviewable, serious work”

A concrete scoring model

Reviewability Score (0–100)

A single number that answers: “Can a human or pipeline meaningfully evaluate this in minutes — and trust what they’re seeing?”

40 — OMC compliance (schema + required fields)
20 — Graph integrity (all references resolve; no orphaned nodes)
15 — Provenance & versioning (identifiers, lineage, revisions)
15 — Narrative packaging (scenes ↔ characters ↔ concepts ↔ shots)
10 — Asset readiness (URLs, metadata, camera/audio fields where relevant)

What “serious” looks like in practice

Dense structure (more typed objects, fewer “blob” descriptions)
Concept packages everywhere (every node gets intent + prompts + references)
Clear lineage (derived-from, who edited what, when)
Low ambiguity (stable IDs, consistent naming, constrained enums)
Fast comprehension (a reviewer can summarize the story + production intent quickly)

The key: “reviewable” is not taste. It’s inspectability, traceability, and hand-off readiness.