News

Is AI Making Volumetric Video Obsolete? The Data Says Otherwise

21/11/2025 – by Natasja Paulssen (CEO/Founder of 4DR Studios)

The conversation keeps coming up in boardrooms and at conferences: with AI advancing so rapidly, do we really still need expensive volumetric capture studios? After all, if algorithms can generate photorealistic faces and reconstruct 3D scenes from single images, surely the days of multi-camera arrays and specialized hardware are numbered.

The European Union’s recent Strategic Research & Innovation Agenda for Virtual Worlds suggests something more surprising. Rather than phasing out volumetric capture, the technology is explicitly identified as a fundamental prerequisite for the next generation of digital experiences. The market seems to agree. Volumetric video revenues are projected to surge from $2.95 billion in 2024 to over $20 billion by 2032, representing a compound annual growth rate exceeding 26 percent.

So what’s really happening? The answer reveals a fundamental misunderstanding about how AI and capture technology actually relate to each other.

The Training Data Paradox

The pattern in the industry reveals something interesting. In January 2024, Meta introduced AI-driven volumetric processing for its Metaverse platforms. Yet by September, Reality Labs was demonstrating Hyperscape, featuring high-fidelity Gaussian Splat captures of physical spaces requiring extensive camera arrays. The company’s product director explained their approach: they’ve tried various AI-only methods, but found greater success with systems that capture real spatial information first, then enhance with algorithms.

This pattern repeats across the sector. Companies with world-leading AI research capabilities continue investing in physical capture infrastructure. The reason becomes clear when you examine how these AI models actually work: they don’t create quality from nothing. Every impressive depth estimation algorithm, every neural rendering breakthrough, has been trained on massive datasets captured in professional volumetric studios.

The EU Strategic Research & Innovation Agenda puts it plainly: “The ability to capture real-world assets, environments, and even people in 3D is a fundamental prerequisite for the realisation of high-fidelity Digital Twins and immersive Virtual Worlds.” The document goes further, noting that while recent AI models like ZoeDepth and Depth Pro perform well on single images, “these models are not yet suitable for dynamic sequences as they lack temporal coherence.”

In other words, the AI that’s supposed to replace volumetric capture needs volumetric capture to exist in the first place.

Where AI Actually Fits

The relationship between artificial intelligence and capture hardware isn’t replacement but multiplication. High-fidelity capture establishes ground truth and generates training datasets. AI then compresses that data, fills gaps between camera views, and optimizes for real-time streaming. Together they enable applications impossible with either technology alone.

Recent research from Frontiers in Signal Processing confirms this hybrid reality. After surveying the landscape of capture technologies, researchers concluded that “RGBD based approaches remain the most popular and widespread technology for capturing sparse camera volumetric video.” While AI-only reconstruction methods show promise, “full dynamic scene reconstruction using multiple camera angles has yet to be demonstrated, and the high GPU processing requirements could make this method expensive.”

The pattern emerging across the industry points toward complementarity rather than competition. Professional capture provides quality ceilings and training foundations. AI provides optimization and accessibility. Markets are differentiating into tiers rather than consolidating into single solutions.

The European Dimension

Europe’s focus on volumetric capture carries additional strategic weight. The SRIA explicitly warns about the continent’s “critical strategic dependency on non-European providers for core XR technologies.” Currently, most advanced capture facilities operate in North America or Asia. This creates not just logistical challenges but questions of digital sovereignty.

When European healthcare systems develop Digital Twins for precision medicine, when museums create interactive historical experiences, when companies build telepresence systems, the training data shaping those AI models matters. Representation matters. Local processing capabilities matter. Control over biometric data definitely matters.

The EU has identified Virtual Worlds as an €800 billion opportunity by 2030, with 860,000 new jobs expected across the continent by 2025. Supporting this growth requires infrastructure, not just algorithms. As the SRIA notes, technologies for real-time 3D capture “enable remote collaboration that genuinely mimics physical presence” and represent research gaps where “no current system seamlessly integrates real-time holography, multimodal feedback, network optimisation, and a human-centred experience.”

Looking Forward

The question isn’t whether AI will transform volumetric video—it already is. Neural compression helps drastically reduce the final volumetric file size. AI-assisted capture workflows are becoming standard. Single-camera solutions are improving rapidly for consumer applications.

But the question of whether AI makes professional capture infrastructure obsolete misunderstands what’s actually happening in the technology stack. The same AI breakthroughs that enable impressive demonstrations from smartphone cameras also increase demand for the high-quality training data that only professional facilities can provide. The same neural rendering techniques that optimize bandwidth also require ground truth captured at higher fidelity than ever before.

Companies betting on AI-only approaches face a ceiling defined by their training data quality. Organizations building on foundations of high-fidelity capture can progressively enhance with AI as algorithms improve, while maintaining quality advantages that compound over time.

The volumetric video market’s robust growth through this period of AI advancement tells its own story. So does the continued investment by major technology companies in physical capture infrastructure. So does the EU’s strategic prioritization of these capabilities in its official research agenda.

AI isn’t threatening volumetric video. It’s changing what volumetric video needs to be and making it more essential than ever.