AR Quick Look vs Scene Viewer: glTF and USDZ Explained

Mobile augmented reality (AR) runtimes allow users to interact with 3D digital objects placed within their physical environments directly from mobile browsers and native applications. Apple’s AR Quick Look and Google’s Scene Viewer are the two dominant native runtimes powering these experiences on iOS and Android, respectively. Rather than relying on heavy, third-party software, these platforms achieve instant rendering and real-world tracking by pairing directly with optimized 3D file formats: USDZ for Apple and glTF/GLB for Google. This article examines how both runtimes leverage these specific formats to optimize performance, visual fidelity, and web-to-AR delivery.

Native Runtimes: AR Quick Look and Scene Viewer

Both Apple and Google developed lightweight viewing pipelines integrated directly into the operating system.

  • AR Quick Look is built into iOS and iPadOS, functioning via Safari, Messages, Mail, and third-party apps. It ties directly into Apple's ARKit and RealityKit frameworks, utilizing the Metal graphics API for hardware-accelerated rendering.
  • Scene Viewer is integrated into Android through Google Play Services for AR (ARCore) and the Google app. It uses Filament, Google’s real-time physically based rendering (PBR) engine, to display assets launched via Chrome or native Android intents.

USDZ: The Backbone of AR Quick Look

Developed in collaboration between Apple and Pixar, USDZ is an uncompressed, read-only ZIP archive that packages Universal Scene Description (USD) files and their accompanying assets (such as textures and audio) into a single payload.

AR Quick Look leverages USDZ in several key ways:

  • Zero-Copy Execution: Because USDZ is uncompressed and memory-mapped, the runtime does not need to extract files to disk before loading. The operating system reads geometry and texture buffers directly from the package, enabling near-instantaneous load times.
  • PBR Materials and Real-World Lighting: USDZ utilizes the UsdPreviewSurface specification. AR Quick Look samples the physical environment using the iOS device camera, calculates ambient lighting and reflection probes, and applies them to the USDZ model's PBR channels (roughness, metallic, and normal maps) in real time.
  • Interactivity and Spatial Audio: Beyond static 3D models, USDZ files can store animation sequences, custom physics interactions, and spatial audio cues. AR Quick Look executes these embedded behaviors natively without requiring JavaScript or custom runtime logic.

glTF and GLB: The Open Standard for Scene Viewer

Maintained by the Khronos Group, glTF (GL Transmission Format)—particularly its binary form, GLB—is often described as the "JPEG of 3D." It packages scene hierarchies, mesh data, animations, and image textures into a compact, binary layout.

Scene Viewer leverages glTF/GLB through the following mechanisms:

  • Compact Binary Delivery: GLB files store vertex buffers, index arrays, and animation matrices in formats that can be uploaded directly to the GPU with minimal preprocessing. This compact footprint is critical for fast transmission over cellular networks before handoff to Scene Viewer.
  • Filament PBR Integration: Google’s Filament engine parses glTF standard PBR materials alongside official Khronos extensions (such as KHR_materials_specular, KHR_materials_clearcoat, and KHR_materials_ior). Scene Viewer maps real-time environmental lighting captured by ARCore onto these material extensions for realistic visual integration.
  • Broad Device Scalability: Because Android encompasses a wide spectrum of hardware capabilities, Scene Viewer relies on glTF’s clean separation of geometry and texture levels of detail (LOD) to maintain target frame rates across varied GPU architectures.

Bridging the File Formats to AR Functionality

Both runtimes interpret their respective formats to enable core augmented reality capabilities:

  • Real-World Scale: glTF and USDZ define internal measurement units (glTF strictly standardizes on meters; USDZ supports defined metric scale tokens). Both runtimes read these scale parameters to ensure an object renders at exact real-world dimensions upon placement.
  • Plane and Surface Detection: The runtime analyzes the physical room using plane detection algorithms, then aligns the model's origin point (defined at the base of the glTF or USDZ mesh) directly flush with horizontal or vertical surfaces.
  • Instant Web-to-AR Handoff: On the web, developers link to USDZ files using standard anchor tags with the rel="ar" attribute in Safari, while Android developers invoke Scene Viewer via explicit intent URLs referencing the remote GLB file. The runtimes bypass complex web page execution to render the file directly within a native AR viewport.