At a glance
○ OpenUSD, MaterialX, and OpenPBR have helped define how an asset should look, and be portable across DCCs and renderers. Delivery hasn't caught up: leaving the workstation still means optimizing for a mesh format like glTF or rendering a turntable.
○ Miris spatial streaming treats 3D like video. The asset is ingested and converted to a stream once, then the 3D data streams adaptively and reconstructs on the viewer's device without consuming a cloud GPU per viewer.
○ Assets originally in the 10s to 100s of GB now stream and render on devices in under a second. More than 30,000 OpenUSD assets have been conditioned by Miris.
○ Evercast, Playbook, and CGAxis use it today for live review, DAM previews, and marketplaces.
Miris at SIGGRAPH 2026. Watch the full presentation here: https://www.youtube.com/watch?v=eCGK6VG3gZY
Inside the studio, the argument is mostly settled. Layout, LookDev, Lighting, and FX pass the same OpenUSD scene description back and forth, and the adoption of MaterialX and OpenPBR means that materials can finally have consistency between renderers. When the asset has to leave the building, most of what we standardized gets left behind.
A supervisor wants to review an asset but has a bad Internet connection. A client wants it on an iPad. A marketplace buyer wants to inspect it before paying. The options are either render a turntable or optimize the mesh and textures so that you can share a portable glTF or USDZ. Most likely both, if you want to support all devices and browsers.
Standards fixed the look, not the delivery
MaterialX provides a portable language for describing materials and shading networks, while OpenPBR builds on that foundation to give renderers a shared, physically-based model for how surfaces should behave and appear. Alongside OpenUSD, these standards are creating a more consistent foundation for 3D content: OpenUSD describes the asset and its scene structure, while MaterialX and OpenPBR describe its materials and appearance. The goal is interoperability across DCCs and renderers without requiring the asset to be rebuilt or its appearance reinterpreted at every step. Anyone who has moved an asset between rendering environments and watched it come out looking different understands the value of that consistency.
But OpenUSD is also an unbounded standard. Production assets run to 100 or 200 GB of geometry, textures, and scene graph. Devices have gotten more powerful and networks faster, but delivering these large assets to the end consumer is still burdensome.
How do you take an asset that strains a workstation, keep the artist's intent intact, and put it on the device in someone's pocket?
Airship fidelity test. Credit: AWS Airship asset, Copyright 2024 Amazon Web Services. All rights reserved.
Treat 3D the way we treat video
People have called OpenUSD the HTML of 3D. The comparison holds, and it points at what's missing. The web became universal when HTML was paired with streaming audio and video. Audio and video started as something you downloaded; now they adapt resolution, bit depth, and sample frequency to whatever device and connection you have, and we expect it to work anywhere.
3D delivery hasn’t been solved in the same way. The standard path is still to process the asset by hand, package it, download it to the client, and run it on the local device’s GPU. Pixel streaming is an alternative that keeps the fidelity by rendering on a cloud GPU and sending video, but every concurrent viewer needs a dedicated cloud GPU and every interaction is a round trip.
Miris spatial streaming conditions the OpenUSD asset once upstream, then streams 3D data that reconstructs on the viewer's device. It doesn't send a decimated file, and it doesn't render frames in the cloud.
The expensive optimization step happens once upstream. After that, delivery and on-device reconstruction scale independently.
No cloud GPU per viewer. The viewer's own device does the reconstruction, so delivery scales with bandwidth, the way video does, not with a fleet of rented GPUs. OpenUSD (and glTF, obj, and stl) goes in. The stream plays in three.js on the web, Unity, and WebXR.
.jpg)
What happens to the asset
We analyze the asset spatially and characterize its material and lighting response. A radiance capture process places digital sensors around and within the asset and records how the asset actually looks, rather than just how it's built. We then build several levels of detail, so the client can pull exactly the data needed given a view, device specs, and network capabilities. The viewer experiences details that refine as they move closer, much like a video player increasing its bitrate.
The Airship test
At SIGGRAPH 2026, we showed the Airship from the Academy Software Foundation's Digital Production Example Library. In its original form it's 44.3 GB, large enough that some of our own workstations struggle to open it. With Miris spatial streaming, it reaches first view in under a second on an iPhone 17 Pro, and it streamed to an iPad over conference Wi-Fi during a presentation on stage.
The Airship isn't a one-off. More than 30,000 assets have gone through the pipeline, exported from Blender, ZBrush, Substance Painter, Cinema 4D, and more. The pattern doesn't change: author in your DCC, export OpenUSD, publish, stream.
Airship test at SIGGRAPH 2026: 44.3GB asset, <1s time to first view, iPhone 17 Pro.
Credit: AWS Airship asset, Copyright 2024 Amazon Web Services. All rights reserved.
Where it's already working
Live review. Evercast integrates with Zoom to support high-fidelity review of 2D and 3D work. With Miris integrated into their platform, each participant gets their own stream of the 3D asset, allowing them to orbit, inspect, and discuss the work independently rather than watching someone else's screen share. This also creates the foundation for annotating and capturing notes directly on the 3D asset, rather than on a static image or proxy.
Asset management. Playbook, a DAM used across media, entertainment, and marketing teams, automatically sends uploaded 3D assets through the Miris pipeline and returns an explorable preview within minutes. We're applying the same approach with the AWS Visual Asset Management System (VAMS) team, where it can support large industrial and digital-twin catalogs.
Marketplaces. CGAxis has uploaded thousands of high-end 3D assets, giving buyers the ability to inspect each asset directly in the browser before purchasing. Instead of relying on static images or simplified previews, buyers can explore the asset with its intended geometry, materials, and lighting.
The same stream reaches desktop browsers, iPhone and Android, Apple Vision Pro, and XREAL glasses, through WebGL, WebGPU, WebXR, or native apps built on the Miris SDK.
Try it on an asset you already have
If your team is exporting optimized glTFs or USDZs for client review, or signing off on turntable renders because the real asset won't open on a laptop, take one of those OpenUSD files and upload it to the free Miris beta. Then open the stream on your phone and compare it to what you're shipping now.
Click here to start streaming in the free Miris beta
About the author
Laurie Koenig is the Senior Director of Product at Miris, a new platform for streaming high-fidelity 3D content. She previously led design engineering at Nike, where she spent more than eight years, and also developed pipelines for DreamWorks Animation.
More about Miris: miris.com