How Spatial Streaming is Unlocking New Possibilities in 3D Asset Delivery

Tell us about your background and how you ended up building delivery infrastructure for 3D.

Sean Looper, Miris CTO: My background is in animation and visual effects. I started in the movie industry as an animator and got drawn into solving visual problems using software. The goal was always the same: how do we use technology to elevate artistry and make the prettiest pictures possible, efficiently and effectively. This was during a period when computer graphics was still nascent but mature enough to be commercially viable, which is an exciting intersection because that's where technology can really shine at scale.

From there, my career spanned infrastructure and deep tech. That journey took me through studios like DreamWorks Animation, where I led a team focused on real-time streaming graphics systems and HPC, and Sony Pictures Imageworks. Over that time period, I worked on more than 60 movies. Then, I led a team at Amazon Web Services (AWS) building a product designed for large-scale, multi-thousand-CPU HPC systems that were purpose-built for real-time processing and streaming of computer graphics.

I began working with Miris about three years ago when artificial intelligence provided the means to address pain points we'd taken for granted as unsolvable physics problems. We looked at the challenge of 3D asset streaming through the lens of three intersecting technologies: streaming graphics, large-scale real-time HPC systems, and AI applied to computer graphics, and it became clear that we could leverage all of them to stream computer graphics at scale without the compromises the industry has accepted for the last 20 years. We realized that these problems were now solvable, that we were uniquely positioned to tackle them with our specialized backgrounds.

Adaptive video streaming has been a solved problem for the better part of two decades. Why has 3D been so much harder to deliver the same way?

Sean Looper: You can experience 3D in 2D, like when we go to the movies, but at the cost of interactivity: I can't change my vantage point or alter the inputs so that what appears on screen responds to me. Game engines offer interactivity, but you’re compromising on visual fidelity for mobility. You can pay for expensive local hardware to improve that experience but then you don't have mobility. You accept trade-offs in exchange for the convenience of having these experiences in your pocket.

In between, you have things like pixel streaming. You can rent a GPU in the cloud, run a game engine there, and view the output in a browser without relying on local hardware, but that introduces new limitations: GPU availability, cost, and latency. You have to be geographically close to that GPU for the best experience. The trade-off here is that you gain portability but at the cost of complexity.

We’ve known and accepted these compromises for decades. We've accepted that real-time graphics require a GPU, and that higher fidelity demands a more expensive setup. We've followed an almost instinctive Moore's Law pattern: better numbers in your system, lower latency, higher bandwidth, more powerful GPUs and CPUs, yield a better experience. We've treated this as an unquestionable law of physics.

We ultimately need to experience 3D as 3D; and AI allowed us to shift where and how we make compromises, pushing much of the computational load into an upstream AI processing phase that, much like LLMs, pre-evaluates the permutations and possibilities within a broader 3D experience. That evaluation becomes significantly less expensive at runtime. Combined with adaptive streaming, we get all the benefits of streaming video without sacrificing the interactivity of a 3D experience. Before AI, that simply wasn't an option.

It was never really a question of "why didn't we stream 3D like video?" but rather “how did we let ourselves assume these were laws of physics rather than problems waiting to be solved with better technology?”

What happens during Miris’ ingest process? Without getting into anything proprietary, what kinds of analysis, transformation, or optimization have to happen before a 3D asset can be streamed efficiently?

Sean Looper: From the customer's perspective, the answer is nothing. That's part of the magic. They take the highest-fidelity representation of their asset, hand it to us, and we take it from there.

Behind the scenes, we run an AI optimization process on that asset. This allows us to distribute it everywhere in the world very cheaply and in a way that lets us pull exactly what's needed, no more, no less, to the end user's device. What we pull takes into account the device's capabilities, current network conditions, the user's point of view relative to the content, and the characteristics of the asset itself.

The overall architecture is straightforward. The asset comes in and a processing step runs for minutes to hours depending on the asset, the settings, and the desired output. Once that's done, the result is effectively a comprehensive spatial database. A smart client then delivers the asset to the viewer based on their device and connection. So outside of that fairly complex optimization step, the system is simple and elegant.

Miris does the expensive work once at ingest, then streams spatial data that reconstructs on the device, rather than rendering frames in the cloud. Walk us through why that distinction matters.

Sean Looper: When you render something, your point of view is fixed – there's no interactivity. Rendering is literally the conversion from 3D to 2D. Game engines tackle this by putting all that data on a GPU and rendering it in real time so you can move the camera around, but at a significant cost to visual fidelity. Pre-rendering, the approach used in visual effects, film, and television, takes the opposite trade-off: it's uncompromising about fidelity but sacrifices interactivity entirely. The end user can't move their view or define their own experience with the content. For movies and TV that's more than acceptable, but there are plenty of applications where that compromise is unacceptable.

Miris provides much of the flexibility of a game engine, in that you can define your own experience and navigate freely around an asset, but at the visual quality you expect from pre-rendered content. You can navigate around a VFX shot or a CG asset at film-quality fidelity without having to view it in some degraded form as you would in a game engine.

What does that actually unlock? What can a team build or ship now that was impractical before, and where does the difference show up: cost, time, or what they can deliver?

Sean Looper: I think there's a lot of answers to that, but the one generating the most interest right now is asset review, which is the ability to share the ground truth of a 3D asset, uncompromised, in any situation, without degrading it just to make it viewable on a phone or across time zones. It's accessibility without sacrificing value, and a lot of our customers and partners are realizing they can have both.

For example, a product designer may work with a globally distributed team. That team may want to review assets from a tablet, an iPhone, or a low-powered laptop with high-fidelity lighting, complex material surfaces, and incredibly high geometric detail intact. These are all things traditional game engines struggle with, especially on low-power devices.

Medical visualization, media and entertainment, e-commerce, and digital twins for industrial use cases are also areas that benefit from spatial streaming. The problem we solve spans dozens of industries because it's fundamentally a logistics problem that has existed for a long time and was simply assumed to be unsolvable.

You're not pitching this as a tool for everyone. Where does solving for 3D delivery create the most leverage, and what do those use cases have in common?

Sean Looper: Our entire team comes from deep tech and high-fidelity computer graphics. We originally set out building a product that we ourselves would want and that's still true. But what we underestimated is that we were also building something that makes 3D easy and approachable for the inexperienced developer who has never worked with 3D before. Previously, developers would have had to spend significant time and effort to condition an asset for a game engine or the web or outsource the work. We've solved that.

Many of our customers already have a high-fidelity asset built by their marketing team, product design group, or engineering team. By taking that asset in its ground-truth form, Miris solves the entire streaming and delivery problem. We underestimated how much relief that would bring.

The result is that 3D has become more accessible and usable for customers who previously either scaled back their usage because of complexity and cost, or never adopted it at all. That's been a real eye-opener for us: we thought we were solving a problem for sophisticated developers already using 3D, but we were simultaneously making 3D approachable for people who had always struggled with it.

Different 3D use cases can have very different requirements, from game assets and digital twins to product configurators, VFX review, and e-commerce. How does Miris adapt to different content types, levels of interactivity, and fidelity requirements?

Sean Looper: Much like Netflix and other streaming services, our expertise is delivering the best possible experience within whatever constraints currently affect the end user – device power, GPU capabilities, network bandwidth, latency, all of it. We're effectively running a continuous optimization algorithm designed to maximize quality in real time.

If conditions are ideal, say you have a 4K HDR display with a fast connection, you get an uncompromising, one-to-one look at your asset. If you're on an airplane viewing it on your phone, you get a slightly reduced version, but one we know you're willing to accept in exchange for the convenience of experiencing that asset wherever you are.

That's the promise we make to our customers: the absolute best experience current technology can deliver, no matter where the end user is or what device they're on.

How does Miris handle scale, whether that means extremely large individual assets, large worlds, many simultaneous users, or frequent asset updates from a live production pipeline?

Sean Looper: Scalability is fundamentally baked into the design of Miris. That means no compromise based on device type, high latency, or unstable connectivity.

On the ingest side, our entire system is designed to scale horizontally. We follow best practices and take seriously our responsibility to meet enterprise-level expectations. Our team's experience at global tech companies brings those instincts directly into how we define and operate the product at scale.

On the delivery side, we can serve assets to millions of end users around the world today. For an enterprise-grade service, we consider that to be table stakes. That's a large part of the value we bring: not just solving this problem, but solving it at enterprise scale, or as we like to say, at internet scale.

For a developer who wants to try this today, what does integration look like, where does Miris fit in an existing pipeline, and where does it not?

Sean Looper: Our SDK is available on GitHub today and is fully supported by us. We also have a Discord channel for developers. On our website (miris.com) and at player.miris.com, you can see examples of our web SDK in action. Getting started is straightforward like with a few lines of TypeScript or even raw HTML.

Our customers typically come to us with an existing 3D engine like Unity, or Three.js on the web and our strategy is to meet them where they are by integrating as natively as possible into those engines. For more specialized deployments like advertising, wearable devices, and other ultra-low-power scenarios, we're also building a lightweight standalone engine designed solely for streaming assets from Miris. Its purpose is portability and minimal memory footprint on devices where efficiency is critical.

We also offer what we call the Playground: a space where web developers can experiment with our API and see a live preview of assets streaming in real time.

Where does spatial streaming go from here? What becomes normal in three years that sounds ambitious today?

Sean Looper: First, we improve 3D where it already lives. 3D is in use all over the place today, many customers, many use cases. We can improve and upgrade those experiences right now in terms of quality, cost-effectiveness, and scale.

Then, we lower the barrier. As the obstacles that previously kept people away from 3D come down, customers, partners, and developers who avoided it or used it sparingly start doing things they've always wanted to do but couldn't. Miris enables that.

Next, is a 3D-first internet. The internet has already gone from image-first (with video as a niche) to video-first. We believe that as 3D becomes as easy to use as video, we'll see a similar shift. Generative AI accelerates this enormously; it lets people who previously needed specialized software skills create 3D content, much the way smartphone cameras democratized video. It used to take a film camera or a Betamax to make video; now it's on everyone's phone.

What is the misconception you most often hear when explaining spatial streaming to developers, and what do you wish more teams understood about the problem Miris is trying to solve?

Sean Looper: Our biggest challenge, and this is normal startup territory, has been getting the word out that these problems are now solved. The pain points no longer exist. Costs are coming down, ease of use is going up, and accessibility is here. It's a paradigm shift, and people need to be informed that the assumptions they've taken for granted are no longer true.

People who have used 3D have felt the pain firsthand and immediately understand the problems we're solving. Our value proposition is obvious and intuitive to them. Others who haven’t wrestled with 3D in the same way may have avoided it because it felt unapproachable. We spend a lot of our time educating the people we talk to that things have changed. Ultimately, they all get it.

For 80 Level’s audience of artists and technical artists, what does spatial streaming change about how they should think about asset creation, optimization, versioning, and delivery? Does it reduce constraints, or simply move those constraints to a different part of the pipeline?

Sean Looper: Spatial streaming removes the burden entirely. I worked in this space for a long time alongside artists and asset creators, and without a doubt, they've always had to consider: how will this asset be used, not just perceptually but technically? What are the limitations of the platform it's destined for? This has been especially true for 3D on the web, where WebGL is effectively a game engine. Assets have to be degraded, beaten up, and optimized for delivery, and that's a burden creators have simply had to carry.

One of the things I love about what we do is that we say, "Make the best possible asset you can. Make it as big as it needs to be, as high-fidelity as it needs to be. Make your textures as large and detailed as they need to be. Same for geometry, same for all of it. Don't worry about delivery. Once you have it, we solve that problem.”

I think it's akin to a filmmaker shooting their movie, the definitive version, the one projected in theaters in all its glory, then uploading it to a service and saying: "Now make this available for Netflix, Hulu, Disney+, YouTube, everywhere else it needs to go. Just solve that for me." That's effectively what we do for 3D. It removes a lot of considerations that most content creators would rather not deal with.

Sean Looper, Miris CTO

Ready to experience spatial streaming? Join the free Miris beta and start experimenting with spatial streaming today. Explore the Playground, and join our Discord community to share work and ask questions. Follow us on LinkedIn and Instagram for the latest news, technical insights, and 3D work from the community.

Evaluating Miris for your organization? Contact us about enterprise deployments and custom integrations.