Fable's splats4D Format Promises Smaller, Dynamic 3D Scenes in Browsers

This review examines Fable co-founder Adam Raudonis's splats4D format, a novel approach to representing dynamic 3D content. We assess its technical claims and implications for real-time web rendering.

For developers working with dynamic 3D content for web or AR, especially those frustrated by the file sizes and processing demands of NeRFs or 3D Gaussian Splatting, Fable's splats4D format offers a compelling alternative. It is not a ready-to-use tool but a foundational technical contribution. Skip if your primary need is static scene representation, where existing Gaussian Splatting solutions are already mature. The bottom line: splats4D represents a significant step towards efficient, streamable dynamic 3D, but its full impact depends on broader adoption and toolchain development.

Methodology

This v0 review draws on the founder's published claims at https://adamraudonis.github.io/splats4D/, accessed on 2026-07-04. Independent benchmarks are pending. Update cadence: re-tested when claims diverge from observed behavior or when a public SDK becomes available. The review covers the technical details of the splats4D format as described by Adam Raudonis, co-founder of Fable, including its architectural differences from 3D Gaussian Splatting and NeRFs. We analyze the founder's assertions regarding file size reduction and real-time browser rendering capabilities, supported by the embedded interactive demos provided in the blog post. What is not covered in this review includes independent performance benchmarks against established 3D formats, long-term workflow integration challenges, or edge cases in complex dynamic scenes. The review does not evaluate Fable's commercial products, only the technical format described in the signal.

A 4D Representation for Dynamic Scenes

Fable's splats4D introduces a novel approach to representing dynamic 3D scenes by extending the concept of 3D Gaussian Splatting into a fourth dimension: time. Unlike traditional 3D Gaussian Splatting, which captures a static scene, splats4D encodes motion and temporal coherence directly within its data structure. Adam Raudonis, co-founder of Fable, describes this as a "4D splat format," where each splat includes temporal parameters alongside spatial ones. This allows for a more compact and efficient representation of changing environments, rather than storing a sequence of independent 3D splat frames.

File Size and Browser Rendering

A core claim of splats4D is its ability to achieve significantly smaller file sizes compared to existing methods for dynamic content. The founder reports that splats4D files are "10x smaller" than a sequence of 3D Gaussian Splatting files for the same dynamic scene. This reduction is attributed to the format's temporal encoding, which avoids redundant data across frames. The blog post features embedded demos showcasing real-time rendering of dynamic scenes directly in a web browser, suggesting the format is optimized for web-based delivery and interactive experiences. These demos run without requiring specialized plugins or heavy client-side processing, indicating a focus on broad accessibility.

Comparison to NeRF and Gaussian Splatting

The splats4D format is positioned as an evolution beyond Neural Radiance Fields (NeRFs) and 3D Gaussian Splatting for dynamic content. NeRFs are known for high quality but are computationally intensive and slow to render, especially for real-time applications. 3D Gaussian Splatting significantly improved rendering speed for static scenes but struggles with dynamic content, often requiring separate splat sets for each frame, leading to large file sizes. splats4D aims to bridge this gap by offering the real-time rendering benefits of splatting while efficiently handling temporal changes, making it suitable for applications like AR, virtual production, and interactive web experiences where dynamic realism is crucial.

What's Interesting / What's Not

The most interesting aspect of splats4D is its explicit handling of the temporal dimension, a critical challenge for photorealistic dynamic 3D content. Current solutions for dynamic scenes, such as storing sequences of 3D Gaussian Splats, are often prohibitively large for web streaming or real-time AR applications. The claim of "10x smaller files" is significant if it holds under independent verification, potentially making dynamic volumetric video a practical reality for a much wider range of use cases. The embedded browser demos, which load and render dynamic scenes interactively, provide tangible evidence of the format's real-time capabilities and web compatibility. This moves beyond pure marketing copy, offering a verifiable experience of the technology in action.

What's less clear is the complexity of authoring splats4D content. While the format promises efficient consumption, the process of capturing and converting real-world dynamic scenes into splats4D is not detailed in the signal. This is a common bottleneck for novel 3D formats; ease of creation is as important as rendering efficiency for broad adoption. The founder's blog post focuses heavily on the technical representation and rendering, but the ecosystem for content creation, including capture systems, editing tools, and conversion pipelines, remains an open question. Without robust tooling, even a technically superior format may struggle to gain traction. Furthermore, while the demos are compelling, the "10x smaller" claim for file size reduction requires rigorous, public benchmarking against diverse dynamic datasets to be fully substantiated.

Pricing

The splats4D format itself is a technical contribution described in a blog post, not a commercial product with a direct price. The source does not detail any commercial offerings from Fable that utilize this format. Any pricing would be associated with tools, services, or platforms built upon this underlying technology. Pricing snapshot date: 2026-07-04

Verdict

Fable's splats4D format is a technically promising development for dynamic 3D content, particularly for web and AR applications where file size and real-time performance are paramount. Its explicit temporal encoding addresses a key limitation of existing splatting techniques for dynamic scenes. For developers and companies building interactive experiences with moving volumetric data, splats4D offers a path to more efficient delivery and rendering. However, its ultimate success will hinge on the development of a robust content creation ecosystem and independent verification of its performance claims. We recommend exploring the provided demos to assess its current capabilities for your specific dynamic scene requirements.

What We'd Test Next

Our next steps would involve rigorous, independent benchmarking of splats4D against a diverse set of dynamic 3D datasets. We would compare file sizes and rendering performance (FPS, latency) against sequences of 3D Gaussian Splats and optimized NeRF implementations. We would also investigate the authoring pipeline: how difficult is it to convert raw volumetric video into splats4D? What are the computational requirements for this conversion? Finally, we would test its robustness with complex, fast-moving scenes and scenes with significant occlusions or topological changes to understand its limitations.

The investor read

The splats4D format signals a critical evolution in the volumetric video and real-time 3D rendering space, moving beyond static scene representations. The market for dynamic 3D content, driven by AR, VR, virtual production, and interactive web experiences, is rapidly expanding. Current solutions are often bottlenecked by file size and rendering performance, making splats4D a potential foundational technology. Comparable tools and formats include Luma AI's NeRFs, Spline, and various 3D Gaussian Splatting implementations. What would make Fable (or a company leveraging this format) highly investable is not just the format itself, but a comprehensive toolchain for capture, authoring, and deployment, coupled with strong adoption metrics. Without a clear productization strategy or an open-source framework that fosters community development, splats4D remains a significant technical contribution that needs commercialization to unlock its full market potential.

Pull quote: “The most interesting aspect of splats4D is its explicit handling of the temporal dimension, a critical challenge for photorealistic dynamic 3D content.”