Volumetric capture records a subject as a real, walk-around-able 3D volume rather than a flat video frame — point a viewer’s phone or a VR headset at the result from a different angle, and they see a different angle of the actual performer, not a reprojected guess. It sounds like it needs a studio full of cameras. For a single-subject capture, it doesn’t: one depth sensor and the right software gets a working pipeline.
The two things a depth sensor actually gives you
A depth camera like the ones used for this work captures two aligned streams at once: an ordinary color image and a depth map — a per-pixel distance reading. Software like Depthkit combines those two streams into a textured 3D mesh for every frame, which is what makes the result a true volumetric recording rather than a flat video with a fake parallax effect layered on top. Depthkit Core, the entry point, is built around exactly this single-sensor setup — Depthkit’s own documentation recommends starting there specifically because it’s “the simplest and fastest way to create volumetric video using a single depth sensor.”
The camera problem nobody’s tutorial mentions
Here’s the catch that trips up a lot of people starting this today: the Azure Kinect DK, the sensor most existing Depthkit tutorials and walkthroughs are built around, has been discontinued by Microsoft. Depthkit’s own documentation still supports it for existing projects, but anyone buying hardware new needs a different sensor. The commonly recommended successor is the Orbbec Femto Bolt, which several volumetric-capture workflows (including Depthkit’s own ecosystem) have migrated to — reportedly with less clock-drift between the color and depth streams than the Kinect DK had, a real practical improvement for anyone syncing multiple sensors. Check Depthkit’s current supported-hardware list before buying anything, since this is exactly the kind of detail that goes stale in a cached tutorial.
A practical first setup
For a first capture: set color resolution to 1080p rather than reaching for the sensor’s maximum — Depthkit’s own guidance calls 1080p “performance friendly and ideal for just getting started,” and a lower resolution means less data to process while you’re still learning the workflow. Leave advanced sensor settings at their defaults for the same reason. Get one clean single-sensor capture working end to end — record, process, view the resulting mesh — before adding a second camera angle or trying to calibrate a multi-sensor rig.
Where this leads
A single-sensor capture already supports a meaningful range of installation work: a performer rendered into a shared AR space, a volumetric loop embedded in a VR environment, or a recorded figure composited into a real-time visual set. Multi-camera setups (Depthkit Studio, built for exactly this) remove the single biggest limitation of one sensor — you can only see the side of the subject facing the camera — at the cost of needing synchronized hardware and real calibration work. Start with one sensor and one subject; the multi-camera rig is a second project, not step two of the first one.
Related Reading
- Introduction to Depthkit — Depthkit Docs
- An Introduction to Depthkit Capture: Depthkit Core, Azure Kinect, and the Refinement Algorithm
- Scatter/Depthkit migrates from Azure Kinect DK to Orbbec Femto Bolt for volumetric capture
- Sparse camera volumetric video applications: a comparison of visual fidelity, user experience, and adaptability — Frontiers in Signal Processing
- Getting Started: Body-Tracking Interactive Installations With Kinect and TouchDesigner