Visionary

Visionary is an open, web-native platform built on WebGPU and ONNX Runtime. Enabling real-time rendering of diverse Gaussian Splatting variants (3DGS, MLP-based 3DGS, 4DGS, Neural Avatars and any future algorithms), and traditional 3d Mesh, directly in the browser. We also support post-processing using feed-forward networks.

Features

Native WebGPU Powered: Utilizes webgpu to achieve high-performance parallel sorting and rendering of millions of Gaussian particles.
Hybrid Rendering Architecture: Automatically handles depth mixing (Depth Compositing) between Gaussian point clouds and standard Meshes, perfectly solving occlusion issues and supporting complex scene compositions.
Universal Asset Loader: Single interface to intelligently identify and load multiple formats:
Static Gaussians: PLY, SPLAT, KSplat, SPZ, SOG
Standard Mesh Models: GLB, GLTF, FBX, OBJ
4DGS/Avatar/scaffold-GS: ONNX
Your own algorithom: See Export your algorithm to ONNX in detail.

Getting Started

Quick Start
Learn how to install dependencies, launch the showcase project (Showcase Scene), and customize model configurations.

Development Guide

1. Foundations

This chapter covers: - Environment Initialization: Configuring ONNX Runtime (ORT) and WebGPU Context. - Universal Loader: Detailed explanation of the loadUnifiedModel interface and intelligent file format detection mechanisms. - Resource Management: Reuse strategies for Gaussian renderers and memory management.

2. Object & Interaction

This chapter covers: - Gaussian Objects (GaussianModel): Property control (Transform, Opacity, Cutoff). - 4D Animation: Timeline control, Play/Pause, and dynamic inference updates.

3. Rendering Pipeline

This chapter covers: - The Hybrid Loop: Understanding the execution order of Scene Pass -> Depth Capture -> Splatting. - GaussianThreeJSRenderer: Detailed API explanation of the core rendering coordinator. - Depth & Occlusion: How Auto Depth Mode works.

Modules Reference

Deep dive into implementation details of specific modules. The documentation is categorized into:

API Reference: Detailed interface definitions, parameter descriptions, and type declarations.
Architecture: System design principles, data flow, and core logic diagrams.

Module Name	Description	Documentation Links
01-App	Application layer entry, context initialization, universal loader	API Reference / Architecture
02-IO	Unified data loading layer, multi-format parsing (PLY/SPLAT/ONNX/GLB, etc.)	API Reference / Architecture
03-Point Cloud	Point cloud data structures, GPU buffer management, data preprocessing	API Reference / Architecture
04-Preprocessing	Data preprocessing, format conversion, optimization algorithms	API Reference / Architecture
05-Sorting	Gaussian particle sorting algorithms, depth sorting optimization	API Reference / Architecture
06-Renderer	Core renderer, pipeline management, GPU resource scheduling	API Reference / Architecture
07-Camera	Camera control, view matrix calculation, projection management	API Reference / Architecture
08-Controls	Interaction controls, orbit controls, user input processing	API Reference / Architecture
09-Shader	WebGPU Shader programs, compute shaders, rendering pipeline	API Reference / Architecture
10-Uniform	Uniform variable management, GPU constant buffers	API Reference / Architecture
11-Utils	Utility library, math calculations, helper methods	API Reference / Architecture
12-Three-Integration	WebGPU renderer integration, hybrid pipeline, rasterization bridging	API Reference / Architecture
13-ONNX	4D dynamic model inference engine, Runtime environment	API Reference / Architecture
14-Managers	Manager pattern implementation, lifecycle management, resource coordination	API Reference / Architecture
15-Timeline	Timeline control, animation playback, frame management	API Reference / Architecture
16-Models	Gaussian model object structure (Splat/Mesh), AABB calculation	API Reference / Architecture
17-Config	Global configuration management, WASM resource path configuration	API Reference / Architecture