Streaming OpenUSD

For decades, carrying a 3D asset between renderers meant accepting that it would come out looking different on the other side. Anyone who has worked in this industry has felt that gap. Today, that gap is closing. MaterialX, open-sourced in 2017, gave the industry a shared way to describe material and shading networks. OpenPBR, introduced in 2023 as a subset of MaterialX, gave it a standard for physically based materials. Together with OpenUSD as the hub, an asset now carries its geometry and scene graph AND a description of what it is supposed to look like.

That is a genuine shift. It also exposes the bigger challenge of delivering that asset to the devices end-users actually have in front of them and in their pocket.

OpenUSD is an unbounded standard. Nothing in it caps asset size, and in production that means assets running to hundreds of gigabytes. Meanwhile 1.7 billion devices were sold last year, with the vast majority being phones. Bridging that distance has meant conditioning: manual optimization, decisions, decimation, export to a lightweight package, download, then render locally as though it were a game asset. Every step in that chain trades away some of the fidelity the standard just established.

At SIGGRAPH 2026, Miris CTO Sean Looper presented a talk where he made the argument that this is the same problem video solved, and it's important to note that video did not solve it by shipping smaller files. It solved it by changing the delivery model. In the talk he draws the comparison directly, treating OpenUSD as something closer to the HTML of 3D: a standard whose value only becomes obvious once a streaming layer sits underneath it.

‍

‍

That is what Miris does. Assets are conditioned once upstream through spatial, material, and lighting analysis, and then stream adaptively, with the client pulling only the data a given device, resolution, and network can use. Frame rate and color accuracy stay fixed, because in 3D they are not negotiable.

Sean demonstrates assets from Blender, ZBrush, Cinema 4D, and Substance Painter being streamed to a browser. For example, he streamed a 44GB airship, an asset that Miris workstations have had trouble opening at all because of the memory it requires, to an iPad over Wi-Fi. He walks through what the pipeline is doing internally, including why none of it involves generative AI. And he closes by streaming to his own phone over 5G from inside the conference center.

Along the way he covers where teams have already deployed Miris in production: live 3D previews inside asset management platforms, full-fidelity assets inside AI training workflows, 3D product views in retail and advertising, collaborative asset review, and across tens of thousands of marketplace assets streamed as previews.

Try Miris

Miris is in open public beta. Upload an OpenUSD or glTF asset and stream it to any device in minutes.