Fable Releases Splat4D Open-Source Format for Streamable 3D Graphics
Fable has launched Splat4D, an open-source, streamable 4D Gaussian splatting format that slashes raw file sizes by up to 58 times for web-native 3D scenes.
- Splat4D achieves file sizes 58 times smaller than raw .splat frames and 14 to 20 times smaller than standard gzip compression.
- The format encodes dynamic 3D scene data at approximately 640 megabytes per second using a Rust and JavaScript decoding engine.
- Web clients can stream and scrub files instantly over standard HTTP Range requests from object stores like Amazon S3 or Cloudflare R2 without custom server logic.
- Pointwise error bounds guarantee absolute precision down to 2 millimeters for spatial position and 4/255 for RGB color channels.
Fable has released .splat4d, an open-source, streamable 4D Gaussian splat format designed to deliver highly compressed, dynamic 3D scenes directly to web browsers. According to technical documentation published by developer Adam Raudonis, the format compresses raw 3D data streams to be 58 times smaller than raw files and 14 to 20 times smaller than gzip compression. A live demonstration showcases a two-second dynamic scene compressed into a single 7.4-megabyte file, which originally required 427 megabytes of raw spatial frames. The library is currently available on PyPI and supports WebGPU-enabled browsers including Chrome 113+, Safari 26+, and Firefox 141+ on Windows.
How does the Splat4D compression engine work?
Splat4D achieves its massive compression ratios by separating static background elements from moving objects. According to the project specifications, the encoder analyzes a 3D sequence and permanently stores static background points only once. In a typical 1.6-gigabyte raw capture sequence, this static background classification reduces the baseline data to just a few megabytes. For dynamic elements, the format utilizes a “deadband hold” tracking mechanism that only records changes when a point’s movement exceeds a user-defined error boundary. This process eliminates quantization flicker and ensures temporal differences remain mostly zero, which dramatically improves downstream compression efficiency.
What are the technical specifications of Splat4D?
The format structures data using an H.265-style closed Group of Pictures (GOP) architecture. It writes absolute keyframes every N frames, followed by P-frames containing exact integer deltas. Because each GOP chunk decodes independently, users can scrub to any point in a timeline without downloading preceding chunks. The file structure consists of a small JSON header carrying bounds and chunk indexes, a static data section, and individual self-contained GOP chunks. Developers can build interactive 3D web applications using these assets, making it an important technology to track alongside the best AI coding tools currently driving web-based graphics pipelines.
How does Splat4D handle streaming and hosting?
Splat4D is built to stream directly from standard, static cloud object stores such as Amazon S3, Google Cloud Storage, or Cloudflare R2. It relies entirely on native HTTP Range requests, removing the need for specialized media servers, manifest files, or complex video containers. To show a complete initial view on screen, a browser client only needs to fetch the first 262,144 bytes containing the header, followed by a single range request for the static background section. When a user scrubs the timeline, the client fetches the prefix of the target GOP chunk to render the first keyframe inside 100 to 150 milliseconds.
Frequently asked questions
What browser engines support Splat4D playback?
Splat4D requires WebGPU support, which is available in Google Chrome 113+, Safari 26+, and Firefox 141+ on Windows operating systems.
What are the default error bounds for Splat4D data?
The default deterministic error bounds are set to 2 millimeters for spatial position, 4/255 levels for RGB color channels, 4/255 levels for opacity, and 2% per axis for scale.
Does Splat4D require a custom streaming server?
No. Splat4D is designed to stream from any static host using native HTTP Range requests. It requires no server-side logic, custom video containers, or manifest files.
If you are looking to build advanced web applications using new 3D graphics formats, check out our tested review of the best AI coding tools.
Best AI Coding Tools (2026): 7 Tested & Ranked →Source: Hacker News. Published July 4, 2026.
The AITW News Desk tracks model releases and AI product launches daily. Every story is fact-checked against its primary source before publishing and edited by Ali Zayed — and always links back to the original source.
AI Tools Worth is independent and unsponsored. Some linked guides contain affiliate links — they never change our verdicts.
Hands-on verdicts, real price changes and the launches that matter. No hype, no spam — unsubscribe anytime.