The challenge
Putting interactive, high-fidelity 3D in front of an on-site user, on whatever device is in their hands, has always forced a tradeoff — and that tradeoff is exactly what makes live markup fall apart. To mark up a model together, a remote expert and an on-site user need the same detailed, interactive model in front of them at the same time. The familiar options cannot deliver all three at once.
Shipping the whole multi-gigabyte model to the device means waiting behind a loading spinner while the asset transfers, and a congested conference or hospital network only makes it worse. To make the model light enough to rotate and annotate at all, teams decimate it, stripping out the fine geometry the expert actually needs to see. The reviewer ends up marking up a stand-in.
The other path — per-session cloud GPU and pixel streaming — can deliver near-exact fidelity. The cost is the infrastructure behind it. Every participant session stands up a dedicated cloud GPU, and the experience is only ever as good as the user's distance from that data center. The farther away they are, the more lag creeps into a workflow that has to feel instant. Either way, there was no practical path to one shared, live, full-fidelity view in front of both people at once.
Why DeviceNexus complements Miris
Miris' adaptive spatial streaming delivers the actual geometry, textures, and materials, rendered locally on the device — with no heavy download, no decimation, and no cloud GPU per viewer. The first frame lands in well under a second, and full detail resolves as the user interacts. That is what separates Miris from both download-first approaches and pixel streaming, and it is what lets the remote expert mark up the real model instead of a proxy.
DeviceView already solved live presence between two people; the next opportunity was to extend that experience by bringing a detailed, interactive 3D model onto the device without the download delays or per-session GPU costs. Miris delivered exactly that — which is why the combination unlocks a workflow neither product could deliver alone.
Reference implementation: DeviceNexus and Miris together
For a reference implementation, DeviceNexus leveraged their partnership with Wacom to build a demo showing the pattern working end-to-end in a healthcare-focused solution. A high-fidelity 3D heart model streamed onto a Wacom MovinkPad — Wacom's new first-to-market enterprise Android tablet — could be inked, rotated, and talked through by a remote expert with the person holding the device, all over conference Wi-Fi.
DeviceView provides the live tele-presence layer and leverages Miris to stream the asset so it can render on-device. In the POC, the integration uses the Miris Web SDK (three.js) plus DeviceNexus-managed Wacom hardware.
Our team was eager to see the results of DeviceNexus and Miris working together to build a tele-presence solution for detail-rich 3D models. Although their initial implementation focuses on a healthcare POC, we see enormous potential for other industries to leverage this capability on Wacom devices.
— Arvind Arumbakkam, Wacom Technology Corp.
A session looks like this:
- Stream the real asset on-device. The Miris Web SDK streams the full-fidelity model and renders it locally. There is no multi-gigabyte download and no decimated proxy, and the first frame lands in under a second.
- Open a live channel. DeviceView opens a secure peer-to-peer WebRTC session that carries voice, spatial data, and markup over one connection. Packets travel screen to screen rather than through a relay, the session starts in under half a second, and the media is AES-256 encrypted.
- Mark up the shared view. The remote expert tele-annotates the model with a DeviceNexus-managed Wacom pen and tablet. Strokes stream across to the on-site tablet in real time.
- Drive the model remotely. The expert rotates, zooms, and pans the model while the on-site user watches it move on their own screen.
- Talk it through. Two-way voice runs over the same channel, so both people discuss the exact view they are looking at.
Results
As the initial healthcare implementation closed, two things stood out. The full-fidelity heart model held up on a constrained conference network — the exact environment where download-first and per-session-GPU approaches struggle most. And standing it up was self-service: DeviceNexus engineering built the POC themselves on the Miris Web SDK, with the asset streamable and running in 30 minutes.
