Definition: Volumetric video is a recording technique that captures people or scenes with many synchronized cameras at the same time and generates a moving 3D representation from them. Unlike a conventional video, you can view the recording from every angle: in VR, in AR, on a tablet or in the browser. Viewers can move freely around the recorded person (six degrees of freedom, 6DoF).
In B2B mechanical engineering, volumetric video mainly preserves the knowledge of real people: experts explain maintenance steps, trainers show manual operations at the machine, and product managers present new products at trade shows and in the showroom.
Advantage: Real people instead of animated figures, with natural gestures, facial expressions and voices.
English: Volumetric Video, Volumetric Capture
German: Volumetrisches Video, volumetrische Aufnahme
How It Works
Recording in the capture studio
- Camera array: Dozens of synchronized cameras surround the capture area
- Simultaneous recording: All cameras record each frame at the same time
- Reconstruction: Software calculates a 3D shape for each frame from the views
- Texturing: The software applies the color information from the camera images to the shape
- Compression: The sequence is compressed for streaming or playback on mobile devices
Result: A sequence of 3D frames that plays like a video and that you can embed in a 3D scene.
Forms of representation
| Form | Principle | Property |
|---|---|---|
| Mesh sequence | One textured polygon mesh per frame | Wide support in engines |
| Point cloud | Colored points in space | Compresses well (MPEG V-PCC) |
| Gaussian splatting | Many small, semi-transparent ellipsoids | Very realistic, new technique |
Gaussian splatting as a current development
In 2023, researchers from Inria presented the "3D Gaussian Splatting" method at SIGGRAPH. It represents scenes as millions of small, semi-transparent ellipsoids and renders them in real time. Hair, fabrics and fine details look much more natural than with conventional meshes.
The extension to moving scenes ("4D Gaussian Splatting") is an active field of research and development. First providers already play back dynamic splats on standalone headsets. For standardization, the Khronos Group published a release candidate of the glTF extension KHR_gaussian_splatting in February 2026, at first for static splats.
B2B Applications
Training: preserve expert knowledge
Scenario: An experienced service technician retires. The company wants to keep his knowledge about the setup of a packaging machine.
Solution: - The studio records the technician volumetrically - The recording is placed next to the 3D model of the machine - Learners see in VR or on a tablet how and where he grips
Advantage: You can follow the manual operations from every perspective.
Service: instructions directly at the machine
Through AR, the recorded expert appears next to the real CNC milling machine and shows how to change a tool. Markerless tracking or marker tracking positions the recording.
Sales and trade shows
A product manager presents a new filling line: as a volumetric figure on the kiosk at the booth, in the virtual showroom or as an AR experience on the tablet of the field sales team.
Design review
Less common, but possible: recorded operators show typical workflows on a new machine design. The team uses this to check reach and lines of sight.
Platform comparison
| Platform | Suitability | Note |
|---|---|---|
| Web | Short clips in the browser | Limit the data volume |
| Mobile/tablet | AR instructions on site | Compression is important |
| Kiosk | Trade show, showroom | Fixed hardware, high quality |
| VR | Training at a 1:1 scale | Full freedom of movement |
Production in Practice
Preparation
A good volumetric recording starts with a clear script. Who speaks, which manual operation is shown, how long can the clip be? Short units of a few minutes are easier to update and to play on mobile devices.
Props and relation to the machine
Large machines do not fit into the studio. The usual method is to record the person with mock-ups or single components and to place the recording next to the 3D model of the machine later. For this, you must agree on scale, position and viewing direction in advance.
Clothing and material
Highly reflective, very fine or plain green fabrics can disturb the reconstruction. Studios usually give recommendations about this.
Post-processing and delivery
- Cleanup: Remove artifacts on hands, hair and edges
- Editing: Shorten and combine sequences
- Output: Generate a suitable quality level for each target platform
- Integration: Embed the clip in the 3D scene, the AR application or the web viewer
- Test: Check the playback on all target devices
Limits and Challenges
Effort and costs
Professional recordings are made in specialized studios with dozens of cameras. The studio landscape is changing: in 2023, Microsoft handed its Mixed Reality Capture Studios over to the provider Arcturus.
Data volumes
Volumetric sequences are much larger than normal videos. Mobile devices and the web need strong compression and short clips.
No interactivity
A recording is fixed. The person does not react to questions. Changes to the procedure mean a new recording.
Lighting
Light and shadows are "baked" into the recording. If the scene does not match, the person looks pasted in. Research on recordings that you can relight later is in progress.
Up-to-dateness
When the machine changes, the video becomes outdated. Recommendation: Use volumetric video for stable procedures, and show changing details with the 3D model.
Frequently Asked Questions (FAQ)
What is the difference between volumetric video and 360° video?
360° video shows the environment from a fixed viewpoint. Volumetric video generates a 3D figure that viewers can move around. Rule of thumb: 360° for environments, volumetric for people.
Do I need my own studio?
Usually not. Companies book productions with specialized studios. Simpler recordings with fewer cameras are possible, but they usually have lower quality.
Does Gaussian splatting replace conventional mesh recording?
Not completely yet. Splats look more realistic, but tools, formats and playback devices are still maturing. Engines and devices support meshes more widely. Trend: Both methods exist in parallel.
Does volumetric video run on standalone headsets?
Yes, with limits on length and level of detail. Compression and streaming determine the quality. Always test recordings on the target hardware.
Which legal points must you consider?
The recording shows a real person. Clarify consent, usage rights and usage period in writing (data privacy). This is especially important when employees leave the company.
Industry Standards and Technologies
- MPEG Immersive Media (ISO/IEC 23090) – series of standards for V3C, V-PCC and MPEG Immersive Video
- 3D Gaussian Splatting (Inria) – original research paper from 2023
- Khronos glTF Gaussian Splatting – glTF extension for splats
- Khronos glTF – open 3D delivery format
- Khronos OpenXR – playback on XR devices
- W3C WebXR Device API – volumetric content in the browser
Related Topics
Capture and reconstruction: - LiDAR – depth capture with a laser - Reverse engineering – 3D from real objects - Sensors – cameras and depth sensors - Photogrammetry – 3D from photos - Gaussian splatting – ellipsoid-based representation - Point cloud – point-based 3D data
3D basics: - 3D model – digital geometry - 3D pipeline – from capture to delivery - Real-time rendering – display in real time - OBJ – simple mesh format - USDZ – AR format for Apple devices
XR technologies: - Virtual reality – training in VR - Augmented reality – instructions at the machine - Mixed reality – combination of real and virtual - Hologram – three-dimensional representation - Immersive experience – immersion in content
B2B applications: - Digital twin – digital replica - WebGL – playback in the browser - Avatar – figure controlled live - 360° video – all-around video from a fixed viewpoint - Remote training – learning without travel
