Explore interactive demos showcasing 3D graphics, physics, AI, and more. Built with cutting-edge web technologies.
Basic THREE.js setup with interactive 3D scene. A starting point for 3D web graphics.
Optimized 3D character model with 45 animations. Demonstrates instanced rendering and animation blending with gltfjsx.
Advanced animation blending with 2D blend spaces, layered animations, and state machines. Features parameter damping, masked layers, and smooth transitions.
First iteration of procedural grass rendering. Features basic wind simulation and LOD optimization.
Enhanced grass rendering with improved shaders. Better performance and more realistic movement.
Advanced instanced mesh rendering for thousands of objects. Demonstrates GPU optimization techniques.
Comprehensive WebGPU materials showcase featuring 10 procedural wood types (40 variations) and 8 industry-standard MaterialX materials. All displayed in a unified grid with interactive controls. Includes chrome, copper, gold, jade, marble, velvet, plastic, and brushed metal.
Click-to-place decals on instanced meshes with BatchedMesh for efficient rendering. Demonstrates raycasting, DecalGeometry, and ring buffer optimization with up to 2000 decals in a single draw call.
Shoot decals and subtract cylindrical holes via CSG. Proof-of-concept that combines surface decals with boolean subtraction on instanced geometry.
WebGPU compute shader rain simulation with 50,000 particles. Features real-time collision detection via render-to-texture, GPU-driven particle physics, and dynamic ripple effects.
Volumetric dust plume with collider streaming into GPU computation textures. Control puff size, turbulence, and scene obstacles while the cyan truck stirs up the debris.
Interactive 3D piñata with physics simulation. Break it open with realistic destruction effects!
Interactive terrain with Rapier physics integration. Demonstrates collision detection on height maps.
Real-time deformable terrain system. Modify the landscape dynamically with physics interactions.
Simplified heightfield implementation for learning. Clean code structure for understanding the basics.
Dynamic crater creation on terrain surfaces. Creates realistic impact deformations with physics.
Bunker demo powered by Rapier physics engine. Fast and accurate physics simulation in the browser.
Breakable wall system with realistic destruction. Uses physics constraints and fracture patterns.
Single-body colliders + Blast Stress Solver with safe-frame splitting and debug lines.
Interactive explosive shockwave with presets (TNT/C4/etc.) and simple scenes.
Spawn humanoid ragdolls with Rapier joints. Click to place them on varied terrain; toggle debug and gravity with Leva.
Navmesh regeneration demo with Rapier physics cubes acting as moving obstacles for crowd agents.
Bunker courtyard rebuilt for Navcat crowds with Rapier debris carving live holes into the navmesh.
Advanced 3D interactive world with physics, multiple AI NPCs, action system, and Gemini chat integration. Complete ECS-style world with HTN planning.
AI-driven NPCs using Hierarchical Task Networks. Watch soldiers plan and execute tactical missions.
Advanced HTN planning with WASM performance. Fluid HTN allows dynamic replanning during execution.
Pure TypeScript HTN planning using htn-ai with the bunker domain. No WASM worker required.
Pure demonstration of Fluid HTN planning system. See how AI agents decompose complex tasks.
Basic AI chat interface with streaming responses. Connect to various LLM providers for conversation.
Enhanced chat with advanced features and context. Includes memory and conversation management.
Interactive NPC conversations with personality. Game-ready dialogue system with character traits.
NPC chat with Rapier physics. Both player and NPCs use character controllers that collide with walls!
Text-based virtual world with AI tool calling. Chat with an AI agent that can move, perform actions, and interact with the environment using real-time tool execution.
OpenAI Whisper speech-to-text in the browser. High-quality transcription running locally.
Text-to-speech synthesis using Kokoro model. Natural-sounding voice generation in real-time.
Voice Activity Detection using Silero VAD. Detect when someone is speaking with low latency.
Real-time voice conversation system. Combines STT, AI, and TTS for natural dialogue.
Smart text chunking for TTS processing. Splits text at natural boundaries for smoother speech.