Futures
Access hundreds of perpetual contracts
TradFi
Gold
One platform for global traditional assets
Options
Hot
Trade European-style vanilla options
Unified Account
Maximize your capital efficiency
Demo Trading
Introduction to Futures Trading
Learn the basics of futures trading
Futures Events
Join events to earn rewards
Demo Trading
Use virtual funds to practice risk-free trading
Launch
CandyDrop
Collect candies to earn airdrops
Launchpool
Quick staking, earn potential new tokens
HODLer Airdrop
Hold GT and get massive airdrops for free
Pre-IPOs
Unlock full access to global stock IPOs
Alpha Points
Trade on-chain assets and earn airdrops
Futures Points
Earn futures points and claim airdrop rewards
Promotions
AI
Gate AI
Your all-in-one conversational AI partner
Gate AI Bot
Use Gate AI directly in your social App
GateClaw
Gate Blue Lobster, ready to go
Gate for AI Agent
AI infrastructure, Gate MCP, Skills, and CLI
Gate Skills Hub
10K+ Skills
From office tasks to trading, the all-in-one skill hub makes AI even more useful.
GateRouter
Smartly choose from 40+ AI models, with 0% extra fees
Fei-Fei Li's World Labs open-sources Spark 2.0, streaming rendering of over 100 million 3D Gaussian points within the browser
ME News report, April 15 (UTC+8), according to 1M AI News monitoring, spatial intelligence company World Labs released the open-source 3D Gaussian Splatting (3DGS) rendering engine Spark 2.0. Its core capability is streaming and rendering over 100 million Gaussian points (splats) to form large-scale 3D scenes in the browser on any device, covering desktop, iOS, Android, and VR headsets.
Spark is built on THREE.js and WebGL2, with the latter being a 3D graphics interface supported by nearly all devices. Most consumer-grade devices can only render 1 million to 5 million Gaussian points at interactive frame rates, while large-scale 3D scans often contain tens of millions to even hundreds of millions of points, with raw data sizes reaching 1GB or more.
Version 2.0 addresses this bottleneck with three new technologies:
Level of Detail (LoD) system: organizes all Gaussian points into a hierarchical tree, where each parent node is a low-resolution approximation of its child nodes. During rendering, it automatically traverses from coarse to fine based on the viewing angle, selecting the optimal subset. By setting a rendering budget (500k to 2.5 million points), it ensures stable frame rates regardless of the total number of points in the scene.
Progressive streaming loading: the newly designed .RAD file format supports random access and progressive transmission. After a scene is opened, a rough outline made up of 64k points is displayed immediately, and then details are loaded step by step with viewing-angle priority. When the user moves, priorities are automatically re-ordered.
Virtual display memory management: allocates a fixed 16 million-point memory pool on the GPU, with 64k-point pages that are automatically swapped in and out. This is similar to an operating system’s virtual memory mechanism, allowing limited display memory to access nearly unlimited scene data.
The core algorithm is written in Rust and compiled into WebAssembly, running in a background Web Worker thread so it does not block the main rendering loop. Spark was originally an internal renderer developed by World Labs for its 3D world generation product Marble; it was later open-sourced as a general-purpose tool.
The release also demonstrated multiple community projects, including a multiplayer space shooter game, Starspeed, built with Marble and Spark. The entire game environment is made up of over 100 million Gaussian points, and the game runs directly through the browser.
(Source: BlockBeats)