The OpenOcean for plant life
You don't download a plant. You describe one — a species, an age, a place, a moment — and it grows. Deterministic. Reproducible. Citable.
The idea
Not a folder of trees. A generative substrate you query for any plant, at any fidelity, at any moment of its life.
Everything downstream — a phone game, a film, a VR world, a climate simulation — asks the same question in a different accent: what would actually grow here, and what would it look like doing so? PhytoEngine is the authority that answers, the way a physics engine answers for gravity. Nobody should ship their own botany.
One plant · every resolution
An N64-detail puppet and a hero-quality photoreal asset are the same object at two zoom levels of one model. You climb continuously — the request just carries how much of the model to evaluate. Drag the slider.
Time is an axis, not a snapshot
The same organism at age 2, 40, 300 — and past it. Growth and decay aren't different assets; they're one trajectory. Run a stand forward and ask what it becomes.
Where reference art enters
Reference photography, approved concept art, digitized specimen photos, player-submitted reference sheets — deposited, versioned and cited, the way a studio keeps a style bible. Every rendered plant can answer why do you look like this? and name the references that shaped it. Disagreement is first-class, not overwritten.
Betula papyrifera · 34 yr · 46.4°N moraine, N-slope · dry Septemberv9.2, seed 0x1F04ref:birch-bark-04 · concept art ref:birch-canopy-v2From one plant to a biome
Many plants competing for light, water and space over decades. Design a fantasy biome or a photoreal one and watch where it stabilizes — or collapses. Same engine, different requester.
Sketch a planting, forecast succession, choose between layouts before a single asset goes into the scene.
Paint an alien-but-plausible forest that obeys competition and spacing, not random scatter — so it reads as a place, not a texture.
Because it's built from cited, inspectable reference material, an art director can study the simulation — and when it diverges from the reference, that's a finding.
Invented species are held to the same standard as real ones — internally consistent growth, silhouette and material — so a fantasy biome still convinces.
The honest question
A far smarter model doesn't need help imagining how a birch grows. It needs the four things raw intelligence can't self-supply — so the smarter the consumer, the more a trustworthy source of ground truth is worth.
No intelligence conjures what a specific reference photo actually shows, or what an art director actually approved. That's a fact of your project — and it lives here.
Regenerating from scratch is plausible but never the same twice, never citable. This returns one addressable, reproducible answer.
Even at 100× speed, a cached, LOD-laddered substrate beats re-solving a canopy from physics every frame, for every viewer.
Confident output still has to be checked against reality. This is where reality is stored.
The place everything that needs a plant goes to ask what it should look like.