Idem datum, multi auctores. same data · many authors · the commons
DREAMWORKS · 2012 · PIXAR · 2016 · ILM · 2018 · ASWF
The studios that compete on every other axis — talent, schedules, awards, opening-weekend revenue — agreed not to compete on data. Three production studios released the formats they had been refining internally for a decade, and a fourth institution agreed to keep them maintained. The data outlasts the tool. The formats outlast the studios that wrote them. This plate is about the institutional move.
For most of computer graphics history, every studio kept its own data formats. An Industrial Light & Magic shot would not open in a DreamWorks pipeline. The geometry was proprietary, the materials were proprietary, the volumes were proprietary, the scene description was proprietary. Sharing a shot between studios meant rewriting half the pipeline. Production data was a moat.
That changed in slow motion across two decades. DreamWorks released OpenVDB in 2012 — its sparse-volumetric format for clouds, smoke, and fire — under MPL. It won an Academy Sci-Tech Award two years later. Pixar released OpenUSD in 2016 — the scene description framework that had run its film pipeline since Coco. ILM, Sony, and Autodesk released MaterialX in 2018 — the shading-network format that lets a look ship between renderers. Three studios, three releases, three formats no one else needed permission to use.
Then, in August 2018, the Linux Foundation and the Academy of Motion Picture Arts and Sciences founded the Academy Software Foundation. The studios agreed not to fight on data. ASWF now hosts OpenVDB, OpenUSD, MaterialX, OpenColorIO, OpenEXR, OpenTimelineIO, OSL, and a half-dozen others — the working infrastructure of the entire industry, vendor-neutral, open-spec, jointly maintained by the studios that wrote them. The film industry built itself a commons.
Three open formats, three studios, three layers of the production data sandwich. Each one was an internal pipeline tool released to the world after a decade of in-house refinement. Read in chronological order, they tell the institutional story of how Hollywood standardized its commons.
SPARSE HIERARCHICALVFX VOLUMES
Ken Museth wrote OpenVDB at DreamWorks for clouds, smoke, fire — the volumetric phenomena that fill any animated film. The trick is that volumes are mostly empty space. A 2048³ grid of fluid simulation has perhaps 0.5% non-zero voxels. OpenVDB's data structure is a sparse, hierarchical tree that stores only the voxels that matter and skips the rest. The format is the algorithm. Houdini, Maya, Blender, Renderman, Arnold, and every modern volume renderer reads VDB natively. Specimen 041.
COMPOSITION ARCSUNIVERSAL SCENE
Pixar built USD inside the studio for the better part of a decade before opening it. Universal Scene Description. The trick is composition: a scene is not a single file but a stack of layers — a base shot, a lighting layer, a shading layer, an FX layer, an animation override — that compose by reference rather than by edit. Pull a layer; the scene re-composes. Every studio above a certain size now lives in USD: ILM, Animal Logic, Weta, Autodesk's Bifrost, NVIDIA Omniverse. The Alliance for OpenUSD (AOUSD) was founded in 2023 — Pixar, Adobe, Apple, Autodesk, NVIDIA — to keep the format open as it became the de facto standard. Specimen 039.
SHADING NETWORKSPORTABLE LOOK
ILM, Sony Imageworks, and Autodesk wrote MaterialX so that a look — the network of shaders, textures, and parameters that make a surface read as wood, copper, skin, snow — could ship between renderers without translation loss. The format describes the network, not its implementation; the renderer interprets. A shot looks the same in Renderman, Arnold, V-Ray, and Karma. MaterialX is now the standard surface description for USD-based pipelines, and the bridge between OSL closures and renderer-specific BSDF graphs. Specimen 040.
Three layers of the production data sandwich. OpenVDB at the bottom — the volumes. MaterialX in the middle — the surfaces. OpenUSD on top — the scene that references all of them. Each format started inside a single studio, was refined under the pressure of feature-film production, and was then released into a commons the entire industry now depends on. Three different studios. One stack.
The formats matter. The institution that holds them matters more. The Academy Software Foundation is the quietest, most consequential thing the film industry has done in the last decade — and almost no one outside the industry knows it exists.
The shape is unusual. ASWF is not a vendor consortium (the vendors are members but do not control it); it is not a standards body (it does not write specifications, it hosts working software); it is not a charity (it operates on roughly $5 million a year in industry dues). It is a shared maintenance institution — a place where the studios collectively pay engineers to keep the working infrastructure working. The board reads like a Hollywood directory: Disney, Sony Pictures, DreamWorks, Universal, Apple, Netflix, Adobe, Autodesk, NVIDIA, Intel, AWS, Microsoft. Every studio that competes on a film line of credit also funds the data layer underneath their pipelines.
What this plate argues is narrow. The right answer is sometimes an institution, not a tool. OpenVDB the format is excellent engineering. OpenUSD the format is excellent engineering. MaterialX the format is excellent engineering. But what makes the three durable — what makes them the working stack of an industry rather than three studios' research projects — is that the studios agreed to share the maintenance. The format is the artifact. The institution is the moat.
Plate IX argued the right answer is often older than the conditions in which it can be used. Plate X argued the right answer is sometimes written by someone the platform hasn't hired yet. Plate XI argued the right answer is sometimes a discipline of restraint at the language layer. This plate argues the right answer is sometimes an institution.
The studios that compete on every other axis — Pixar, DreamWorks, ILM, Sony, Disney, Netflix, Apple — agreed not to compete on data. They released the formats that ran their pipelines, and they jointly funded an institution to keep those formats maintained. The film industry, which is otherwise as proprietary an industry as exists, built itself a working commons. The commons is not a slogan. It is a budget line, a board, a release cadence, a CI pipeline.
The discipline of formats is the discipline of trusting the data. Don't ship the tool that wrote the data; ship the data with a spec the next tool can read. Don't keep the format proprietary; release it and let the institution maintain it. Idem datum, multi auctores. Same data, many authors. The format outlasts the studio that wrote it. The institution is the reason the format outlasts.