Hololight and SpectreXR Combine XR Streaming with Physics-True Hand Interaction
Two limits have shaped enterprise XR for years. Engineering models grew heavier than headsets could carry. Interaction never got much further than pointing at things.
Hololight and SpectreXR are combining their technologies to remove both at once. Hololight's XR streaming infrastructure renders the full application on server or workstation hardware and streams only pixels to the headset, so assemblies arrive at engineering density without being cut down first. SpectreXR's OctoXR resolves what happens when a hand closes around them: ten fingers, real contact, real weight, objects that stay where they are put.
For enterprises in aerospace, defense, automotive, and manufacturing, that combination makes something available that has not been until now. The real asset, handled the real way.
"Everyone already knows how to pick something up. That's thirty years of learning, and it should transfer straight into a headset. Hololight put the real model in front of the person. We make it something they can take hold of. For the work these customers do, the asset and the hands are one problem, not two."
– Ivan Rajković, Founder and CEO, SpectreXR
The division of work is clean, and each side does what it is built for.
1. Hololight streams the rendering. The application runs on cloud, on-premises, or air-gapped infrastructure. The headset acts as a display. Complex models reach the device at full fidelity, no data preparation, no polygon reduction, and no sensitive data stored on the end device.
2. OctoXR resolves the interaction. OctoXR is an interaction layer, not a tracking system. It takes the bone-level hand pose the device already provides and turns it into physics-true grasping, two-handed manipulation, spatial UI, and configurable gestures, with one interaction model across the devices an enterprise actually deploys.
Nobody has to be taught this part. A technician entering a session already knows how to pick something up, turn it, pass it between hands, set it down. That knowledge arrives intact, with no training on the tool before training on the task.
"What impressed me about OctoXR is how far it goes beyond gesture recognition: real physics, real contact points, real weight, not a few recognized gestures layered on a scene. And it works with a fully rendered, engineering-density asset, not a simplified one built to fit on the device, exactly the missing piece for enterprise XR's interaction side."
– Alex Werlberger, CTO, Hololight
The two layers are not simply stacked. Moving the rendering off the device frees the compute that interaction quality actually runs on.
Stable grip. Ten fingers on one object means dozens of contact points to resolve every frame. How well that resolves depends on how much solver time is available. Too little, and grips creep, objects buzz against the palm, and contact drops as the scene fills up. With rendering offloaded, there is enough budget to settle it properly.
Fine detail. Collision geometry can stay close to the real mesh instead of being reduced to rough shapes. Latches, threads, detents, and tight fits are exactly the parts of a procedure people have to practise.
Whole assemblies. Everything simulates at once, so a mechanism can be worked in place rather than as an isolated part.
Speed and weight. Fast placement and catching resolve properly instead of passing through, and higher contact resolution lets inertia behave honestly.
What offloading does not solve is latency, which is why the input path stays light. Interaction data is cheap to move. Pixels are not.
"Contact resolution is the whole game. Give it enough budget to converge and the object gets weight, friction, and stillness. Everything that makes a hand believe it."
– Matija Fumić, Lead Software Engineer, SpectreXR
For training, the difference is between a session that demonstrates a procedure and one that teaches it. When the object behaves, stays still in a closed hand, pushes back where it should, and passes cleanly between hands, the trainee is performing the actual motion, the same grip, the same reach, the same commit on a switch or a latch. Repetition under those conditions builds motor memory, and the skill leaves the headset with the person.
For engineering, it is a part that can be picked up and understood. How it seats. Where it fouls. Whether two hands can reach it at all. Those are answers that come from handling something, not from rotating it on a screen.
Because the detail survives the pipeline, the hands can find it. A latch catches. A bolt threads. A switch clicks.
OctoXR is an interaction layer, not a tracking system. It consumes bone-level hand pose from whatever tracking the device already provides and resolves it into physics-true grasping, two-handed manipulation, spatial UI, and configurable gestures. One interaction model across Unity, native visionOS, OpenXR devices, and streamed XR. Training teams use it so trainees practise the real motion. Design teams use it to handle parts before they exist. Device makers and software platforms license it to embed. Built by SpectreXR, founded in 2016 and focused on hand interaction since 2021.
Hololight runs any application on any major AR or VR device, instantly and without compromise. We built the industry's leading XR streaming infrastructure for enterprises, the only complete proprietary pixel-streaming stack on the market, deployable to cloud, on-premises, and air-gapped environments.
We radically decouple software from hardware, making Hololight the strategic choice for global scalability. By streaming pixels instead of data, we ensure uncompromised data security and unmatched performance. Our orchestration platform brings a fragmented XR landscape into one secure and scalable IT layer, unrestricted by any single device or vendor.