The product

A pipeline, not a prompt.

Python owns the logic and the guardrails; the Nova engine collaborates within them. Every stage has a deterministic gate that rejects anything that fails — so what ships is correct, not just plausible.

The canonical scene-spec

Every explainer is first a versioned scene-spec — subject + pedagogy plan + visual scene + narration script + timing manifest, in one JSON. Every output format renders from it. That is what makes localization, re-branding, re-rendering, and an editable-source premium tier possible — and it gives the guardrails a schema to validate before a single frame is drawn.

All possible outputs

Author once as a scene-spec; ship it as anything.

Interactive HTML5

LIVE

self-contained, drag/play/scrub

Embed / Web Component

ROADMAP

paste into any CMS or app

MP4 + captions

ROADMAP

guaranteed playback anywhere

GLB / USDZ

ROADMAP

3D + AR via model-viewer

SCORM · xAPI · LTI 1.3

ROADMAP

drops into any LMS with tracking

Scene-spec JSON

ROADMAP

editable source — localize & re-brand

The bar we build to

World-class, measured.

Benchmarked against Brilliant, 3Blue1Brown/Manim, PhET, Desmos & Wolfram.

✓ Motion with intent — every transition explains something

✓ Sub-frame narration sync — voice, captions, highlights locked to one timeline

✓ WCAG 2.2 AA — screen-reader descriptions, keyboard, reduced-motion, captions

✓ Responsive — phone to 4K, touch + mouse + keyboard

✓ <100 ms direct manipulation — drag/slider recompute live

✓ One concept, progressive disclosure, immediate feedback

✓ 60 fps budget with graceful video fallback

✓ Localization-ready — separable text + per-language audio

✓ Deterministic, reproducible from a versioned scene-spec