1. What Is a Spatial Web Browser?

A Spatial Web Browser is a user agent that loads, interprets, and presents Web content (HTML, CSS, JS, WebGL/WebGPU, WebXR, media) directly inside a 3D coordinate space instead of flattening everything onto a 2D rectangular viewport. Every DOM element (text nodes, images, canvas, form controls, SVG, etc.) can be:

• Positioned, rotated, and scaled in world / XR reference spaces
• Layered with true depth ordering (not just z-index compositing) for stereoscopic correctness
• Interacted with using spatial input sources (gaze, hands, controllers, future: eye tracking, anchors)
• Composed alongside native 3D assets (GLTF models, environment maps) in one unified frame loop

In short: a Spatial Web Browser lets "regular Web pages" become immersive 3D experiences without abandoning open Web standards.

2. Why Not Just Extend an Existing (Classic) Browser?

Traditional engines (Blink, Gecko, WebKit) are extraordinarily capable—but architecturally optimized for 2D document + compositor pipelines. Retrofitting full 3D spatial semantics collides with deep assumptions:

1. XR Spatial Integration Requirements: In XR systems, we need to seamlessly integrate Web content with other Unity/Unreal 3D engine content in spatial environments, which traditional browsers' 2D architecture cannot satisfy.
2. Full 3D Element Design: Inspired by visionOS, in Spatial Web, we need all elements to have true 3D spatial properties rather than being confined to a flat plane, requiring a complete architectural redesign from the ground up.

Therefore JSAR pursues a purpose-built Spatial Web Browser Engine rather than modifying an existing one.

This approach contrasts with the evolutionary path taken by visionOS Safari, which gradually introduces spatial features like HTMLModelElement for 3D model display and Spatial Photo support for stereoscopic images. While visionOS represents a progressive enhancement of traditional browser engines toward spatial capabilities, JSAR chose a fundamentally different direction—designing for spatial (3D) environments from the ground up rather than retrofitting existing 2D architectures.

Both approaches ultimately converge toward the same goal: enabling web developers to create true 3D applications using familiar web technologies without learning new frameworks or specialized knowledge. Whether through gradual evolution or ground-up redesign, the future of spatial web development will remain accessible to existing web developers while unlocking the full potential of three-dimensional user interfaces.

3. Definition: Spatial Web Browser Engine

A Spatial Web Browser Engine is the runnable core (parsing, layout, styling, scripting, rendering, device integration) specialized for spatial presentation. In JSAR this means:

Standards Alignment

HTML5 subset, CSS3 partial, DOM APIs, WebGL1/2, WebXR spaces & inputs.

Spatialized DOM

Each element resolves to a layout object enriched with layer + transform metadata.

Unified Render Passes

Discrete passes (onBeforeRendering, onOpaquesRenderPass, onTransparentsRenderPass, onAfterRendering) integrate seamlessly with host 3D engines.

Efficient Batching

Aggressive draw-call minimization for stereoscopic + high-refresh headsets.

Embeddability

Loaders (Unity available, Unreal planned) treat the engine as a library.

Developer Tooling

Chrome DevTools Protocol (Runtime domain) with WebSocket server implementation to enable Chrome DevTools and other CDP clients for web application debugging. Future plans include rendering debugging tools spatially for enhanced visualization capabilities in specific use cases.

4. Core Capabilities

4.1 Spatialized DOM

Every HTML element carries intrinsic 3D spatial properties, not just 2D positioning.

• CSS transform functions work natively in 3D space (translate3d(), rotateX/Y/Z(), scale3d())
• Layer field metadata per LayoutObject for depth ordering
• Elements can be positioned using spatial units conceptually (meters, not just pixels)

Example:

.spatial-panel {
  transform: translate3d(2px, 1.5px, -3px) rotateY(15deg);
  /* Panel positioned 2px right, 1.5px up, 3px forward, rotated 15° */
}

4.2 Unified Graphics Pipeline

HTML elements and 3D content can be seamlessly integrated in JSAR, enabling powerful combinations like HTML + Three.js for mixed reality applications. HTML serves as the foundation for implementing Spatialized GUI components, while Three.js handles 3D model rendering and complex 3D applications within the XR space.

JSAR's WebGL/WebXR Foundation:

JSAR achieves this integration through comprehensive support for WebGL and WebXR, allowing popular frameworks like Three.js, Babylon.js, or custom web rendering engines to run natively on the platform.

Key Difference in Initialization:

Unlike traditional browsers where you request a canvas element and draw scenes onto it, XR spatial environments don't require a canvas. Instead, JSAR provides direct access to navigator.gl - a WebGL2 context that allows you to draw objects directly into the spatial environment.

Example: A virtual museum implementation demonstrating HTML + Three.js integration:

<html>

<head>
  <meta charset="utf-8" />
  <title>Virtual Museum - JSAR Spatial Browser</title>
  <script type="importmap">
    {
      "imports": {
        "three": "https://ar.rokidcdn.com/web-assets/yodaos-jsar/dist/three/build/three.module.js"
      }
    }
  </script>
  <style>
    .info-panel {
      position: absolute;
      background: rgba(0, 0, 0, 0.8);
      color: white;
      padding: 20px;
      border-radius: 10px;
      width: 300px;
      transform: translateZ(100px);
    }

    .artifact-info {
      left: 50px;
      top: 100px;
    }

    .controls {
      right: 50px;
      top: 100px;
    }
  </style>
</head>

<body>
  <!-- HTML UI Panels for Interactive Information -->
  <div class="info-panel artifact-info">
    <h3>Ancient Vase</h3>
    <p>Ming Dynasty ceramic vase from 1368-1644 CE</p>
  </div>

  <div class="info-panel controls">
    <h3>Museum Controls</h3>
  </div>

  <script type="module">
    import * as THREE from 'three';

    let scene, camera, renderer;
    let artifactMesh;

    async function init() {
      // Initialize Three.js with navigator.gl (no canvas needed)
      const gl = navigator.gl;

      try {
        scene = new THREE.Scene();
        camera = new THREE.PerspectiveCamera(75, window.innerWidth / window.innerHeight, 0.1, 1000);
        renderer = new THREE.WebGLRenderer({
          canvas: { addEventListener() { } },
          context: gl,
        });

        // Create 3D artifact (vase)
        const geometry = new THREE.BoxGeometry(0.5);
        const material = new THREE.MeshPhongMaterial({ color: 0x8B4513 });
        artifactMesh = new THREE.Mesh(geometry, material);
        artifactMesh.position.set(0, 0, 0);
        scene.add(artifactMesh);

        // Fit the scene to the proper size
        function fitTo(scene, targetSize = 1) {
          const box = new THREE.Box3();
          scene.traverse(object => {
            if (object instanceof THREE.Mesh || object instanceof THREE.Group) {
              box.expandByObject(object);
            }
          });
          const size = box.getSize(new THREE.Vector3());
          const scale = targetSize / Math.max(size.x, size.y, size.z);
          scene.scale.set(scale, scale, scale);
        }
        fitTo(scene, 0.2);

        // Add lighting
        const ambientLight = new THREE.AmbientLight(0x404040, 0.6);
        const directionalLight = new THREE.DirectionalLight(0xffffff, 0.8);
        directionalLight.position.set(1, 1, 1);
        scene.add(ambientLight);
        scene.add(directionalLight);

        // Set up XR session
        const session = await navigator.xr.requestSession('immersive-ar', {});
        const baseLayer = new XRWebGLLayer(session, gl);
        session.updateRenderState({ baseLayer });

        renderer.xr.enabled = true;
        renderer.xr.setReferenceSpaceType('local');
        renderer.xr.setSession(session);

        // Start render loop
        renderer.setAnimationLoop(animate);
      } catch (err) {
        console.error('Failed to initialize XR:', err);
      }
    }

    function animate() {
      artifactMesh.rotation.y += 0.01;
      renderer.render(scene, camera);
    }

    init();
  </script>
</body>

</html>

This example demonstrates how HTML <div> panels provide interactive UI controls and Three.js renders 3D artifacts using navigator.gl instead of canvas - both coexisting in the same 3D scene with proper depth ordering and spatial relationships.

4.3 Standards Compatibility

Developers can reuse existing Web knowledge instead of learning proprietary 3D APIs.

• HTML5 subset: <canvas>, <img>, <audio>, form elements, semantic sections
• CSS3 partial: transforms, gradients, grid layout, text styling
• JavaScript APIs: DOM manipulation, WebGL1/2, WebXR Device API
• Modern standards: ECMAScript modules, Web Workers, WebAssembly

Example: Existing web developers can create spatial UIs using familiar <div>, CSS Grid, and JavaScript without learning Unity or Unreal Engine-specific APIs.

4.4 XR & 3D Integration

Native support for immersive viewing modes and spatial interaction paradigms.

4.4.1 Stereo Rendering

JSAR supports both mono and stereo rendering modes to provide immersive experiences across different display types and platforms:

Native Binocular Devices: On devices with natural dual-eye displays (VR/AR headsets), JSAR enforces stereo mode for dual-eye rendering. All Web Content is automatically rendered from both left and right eye perspectives, creating true stereoscopic 3D experiences.

Desktop Platforms: Users can choose between mono and stereo rendering modes for Web content. This flexibility allows developers to test stereo experiences on traditional displays or provide optional 3D viewing modes.

4.4.2 Spatial Audio System

JSAR provides automatic 3D audio spatialization for all audio elements without requiring complex Web Audio API setup, making spatial audio accessible to all web developers.

Every <audio> element automatically gains spatial properties based on its 3D transform position, with distance attenuation, directional audio, and environmental acoustics applied automatically.

Example:

<audio src="ambient.ogg" autoplay loop></audio>
<!-- Audio automatically plays from the left side, 2 meters up -->

When users' positions change relative to the page in space (whether due to user movement or HTML panel movement), they can perceive the distance and position changes of sound sources through audio cues. This dynamic spatial audio positioning provides users with natural auditory feedback, enhancing the sense of immersion and directionality in spatial environments.

4.4.3 Spatial Images

JSAR supports stereoscopic images that provide true depth perception in XR environments through the spatial="stereo" attribute extension. JSAR handles the proper rendering to each eye for depth effects, supporting side-by-side stereo format.

Example:

<img src="stereo-photo.jpg" spatial="stereo" />
<!-- Displays with proper depth separation in VR -->

4.4.4 WebXR Input Sources

JSAR internally leverages WebXR Input Sources API to bridge traditional DOM events with XR device interactions, supporting various input methods including gaze tracking, hand tracking, and motion controllers.

By mapping XR device interactions to native DOM events (such as mouseup, mousemove, click), JSAR allows web developers to use familiar DOM event listeners to respond to gestures, controller rays, touchpads, and other XR inputs without needing to write additional WebXR Input Source code.

Example:

// Standard WebXR input works with spatial elements
document.querySelector('.spatial-button').addEventListener('click', (event) => {
  // Triggered by gaze, hand tracking, or controller input
  console.log('Spatial button clicked:', event);
});

4.5 Performance Batching

Design Philosophy: Inherently 3D HTML Elements

JSAR makes every HTML element inherently 3D without requiring explicit user declarations, driven by three core principles:

1. Content-Focused Development: We enable users and AI assistants like Copilot to write HTML based on what they want to express, rather than being constrained by performance considerations. Developers can focus on semantic content without worrying about spatial rendering overhead.

2. Complete Platform Backward Compatibility: JSAR provides a seamless cross-platform development experience without requiring additional XR-specific declarations. Existing web content works naturally in spatial environments, ensuring developers aren't limited to XR-only applications.

3. Spatial Reality Assumption: In spatial contexts, any 2D element should be fundamentally 3D by nature. Based on this assumption, JSAR implements aggressive optimization using instanced draw calls to minimize GPU overhead while maintaining rich spatial GUI capabilities.

JSAR achieves exceptional batching performance because all HTML elements—whether <div>, <p>, <img>, or even text nodes—are fundamentally rendered as textured quads in 3D space. Since every element shares the same underlying geometric primitive (a quad), they can be efficiently batched together with just different positions, sizes, and textures/materials. This unified rendering approach allows thousands of diverse DOM elements to be drawn in a single GPU call.

4.6 Developer Experience

Familiar debugging tools work in spatial environments.

• Basic Chrome DevTools Protocol (CDP) WebSocket service and communication protocol
• Foundation for real-time debugging infrastructure
• General Domain support coming soon once the system stabilizes

4.7 Extensibility

The browser engine embeds as a library into existing 3D engines and workflows. Taking Unity as an example, JSAR integrates seamlessly into Unity's rendering pipeline by inserting Web content CommandBuffers at different stages of both Built-in and Universal Render Pipeline (URP).

During the Opaque rendering stage, Unity and Web content are rendered to the same framebuffer, enabling correct depth calculations between Unity objects and HTML elements. This ensures that a Unity cube can properly occlude a Web-based UI panel, or vice versa, based on their actual 3D positions.

Similarly, in the Transparent rendering stage, JSAR maintains the same framebuffer sharing approach, allowing transparent Web elements to correctly blend with Unity's transparent materials while preserving proper depth sorting and alpha blending operations. However, transparent object rendering differs slightly from opaque object rendering, as it typically relies on object rendering order to achieve proper transparency blending effects. Therefore, when JSAR's transparent objects blend with Unity's transparent objects, some edge cases may occur. This issue will be resolved in the future by introducing a unified OIT (Order-Independent Transparency) algorithm.

JSAR currently supports only Unity SDK integration. However, as described above, developers can also use the JSAR native library to insert Web Content CommandBuffers into other engines' pipelines to achieve mixed rendering. And the JSAR library currently works on macOS and Android platforms, supporting OpenGL and GLES 3 rendering backends.

5. Architectural Comparison (Classic vs Spatial)

5.1 Primary Surface Differences

Classic Browser: The fundamental rendering target is a 2D rectangular viewport that gets composited to the screen. All content is ultimately flattened onto this 2D surface, even when CSS 3D transforms are applied.

Spatial Browser: The primary surface is a 3D world coordinate system or WebXR reference spaces. Content exists natively in 3D space with true volumetric positioning, not flattened projections.

Impact: This architectural difference enables true spatial interfaces where UI elements can exist behind, in front of, or alongside 3D objects with proper occlusion and depth relationships.

5.2 Hit Testing Methods

Classic Browser: Uses 2D box model calculations with stacking context ordering. Ray casting is simulated by projecting 3D transforms back to 2D coordinates.

Spatial Browser: Implements true ray/volume intersection testing with depth-aware ordering. Spatial elements can be tested for intersection in 3D space using their actual geometric bounds.

Impact: Enables natural spatial interaction paradigms like gaze-based selection, hand tracking, and controller pointing with accurate depth perception.

5.3 Input Source Architecture

Classic Browser: Designed around mouse, touch, and keyboard input. XR inputs are emulated by synthesizing mouse events from controller positions.

Spatial Browser: Native support for spatial input sources including gaze rays, articulated hand tracking, 6-DOF controllers, and future technologies like eye tracking and spatial anchors.

Impact: Enables natural multimodal interaction in spatial environments without the limitations of mouse event emulation.

5.4 Embedding Model

Classic Browser: Operates as a system-level application creating windows or tabs. Designed to be the primary application rather than a component.

Spatial Browser: Functions as a library that embeds within host 3D engines (Unity, Unreal, custom). The spatial browser becomes a component within larger 3D applications.

Impact: Allows existing 3D applications and games to incorporate web content as spatial UI elements or information panels.

5.5 Pipeline Control Granularity

Classic Browser: Provides limited hooks into the rendering pipeline. Developers can influence layout and styling but have minimal control over the actual rendering passes.

Spatial Browser: Exposes explicit pass lifecycle APIs (onBeforeRendering, onOpaquesRenderPass, onTransparentsRenderPass, onAfterRendering) allowing fine-grained integration with host rendering systems.

Impact: Enables advanced rendering techniques like custom depth sorting, multi-pass effects, and integration with existing 3D engine rendering pipelines.

6. Developer Experience & Use Cases

6.1 For Traditional Web Developers

If you're a traditional web developer, you can seamlessly spatialize your existing web applications without learning new languages and frameworks. Your familiar HTML structure, CSS layouts, and JavaScript logic work directly in 3D space:

• Standard DOM manipulation and event handling remain unchanged
• Progressive enhancement: add spatial transforms via CSS without breaking 2D fallbacks
• Your web app runs identically on both traditional browsers and spatial environments

6.2 For WebXR Developers

If you're already developing WebXR applications, JSAR offers significant advantages:

• Direct spatial execution: Run your WebXR apps natively in spatial environments without browser overhead
• Familiar GUI toolkit: Continue using HTML + CSS for spatial UI instead of learning 3D GUI frameworks
• Seamless integration: Combine WebGL 3D content with HTML panels using standard web APIs
• Enhanced input handling: WebXR input sources automatically map to familiar DOM events

6.3 For Desktop Innovation

Spatial Web Browser isn't limited to XR devices. On desktop platforms, it enables entirely new ways to interact with web applications:

• 3D Infinite Canvas: Arrange multiple web pages in 3D space like a limitless desktop
• Spatial multitasking: Position different web apps at various depths and angles for optimal workflow
• Immersive data visualization: Transform traditional dashboards into explorable 3D environments
• Novel interaction patterns: Use mouse and keyboard to navigate through spatial web content

7. Conclusion

JSAR represents a fundamental paradigm shift in web browsing, transforming the traditional 2D document model into spatial browsing experiences across all platforms. Supporting XR devices and desktop platforms (with future consideration for mobile platforms), JSAR bridges the gap between familiar web development practices and immersive spatial web browsing by reimagining every HTML element as an inherently 3D entity.

The engine's architectural innovations—from spatialized DOM with true depth ordering to unified graphics pipelines that seamlessly blend HTML and WebGL content—demonstrate that spatial web applications need not abandon web standards. Through comprehensive WebXR integration, automatic spatial audio positioning, and intelligent input source abstraction, JSAR enables developers to create immersive experiences using the same HTML, CSS, and JavaScript skills they already possess.

JSAR's embedding model as a library within existing 3D engines like Unity, combined with its aggressive batching optimizations and cross-platform support (macOS/Android with OpenGL/GLES3), positions it as a practical solution for both XR applications and desktop innovation. Whether enabling traditional web developers to spatialize existing applications, empowering WebXR developers with familiar GUI toolkits, or unlocking novel interaction patterns like 3D infinite canvas on desktop platforms, JSAR makes spatial computing accessible without sacrificing the open, standards-based nature of the web.

By choosing ground-up architectural design over retrofitting existing browser engines, JSAR achieves the clarity and performance necessary for the next generation of spatial web experiences, while maintaining the developer accessibility and cross-platform compatibility that make the web universal.