🚀 Major upgrades just landed in SuperSplat, the free and open source platform for 3D Gaussian splatting. Here is a 24-MILLION-Gaussian scan streaming live to a web browser. Near instant load time. Solid 60 fps. How? A new compute-based WebGPU renderer + automatic LOD streaming. 🧵

Finally shipped my russian roulette game 🔫 Pixel-art, 5 players, 1 bullet each. The multiplier climbs every second, but anyone who cashes in resets it for the whole table. First to $500 wins. Solo vs bots, or send friends a link: russianroulette.online

Spent some time yesterday updating the #SuperSplat viewer: * Added a controller 🕹️ info popup * Updated annotations look and feel * Improved the way we handle cameras embedded in collision (no longer stall for voxel download cc @fulligin). "Paul Kirk Japanese House" by garrettknelli superspl.at/scene/7f0a5157 Viewer source github.com/playcanvas/supers…

You can now run the @PlayCanvas Engine headlessly in Node.js without JSDOM! Here's how: developer.playcanvas.com/use…

The @playcanvas Editor 2.32 is out ✨ 🔗 Build physics contraptions with the new Joint component ⛰️ Add real surface depth with parallax occlusion 🌓 Pick which shadow cascades each object uses 📐 Justify wrapped text 🎯 Reworked Gaussian splat LOD controls

A pivotal moment for real estate! 🏡 Capturing a property in 3D used to mean a $5,000 LiDAR scanner. This house was scanned with a $500 @insta360 X5 - a consumer 360 camera. And best of all? All of the software is free and open source! It's based on a technique called 3D Gaussian splatting: 🪄 Splat creation: Spirula Studio 🖼️ Splat rendering: @PlayCanvas So for $500, you can present any property online and let prospective purchasers freely explore as if it was a videogame. 🎮 [1/3]

Replaced splat-transform's Gaussian scale DOF approximation with multi-frame accumulation. This f/0.8 shot rendered 256 samples in 8s on my Mac M4. 🔗 superspl.at/scene/7e4e9bcb

I filmed a short walk through an interior unfinished temple and let K3 train it into a 3D model on website for me. Pipeline is basically: video to sharp frames to camera poses to splat training to a compressed web model. Then PlayCanvas renders it in the browser. No special backend. Just static files. Wild part is how ordinary the capture felt. Phone video in, navigable room out. And the model traininig will be more define when you can take as many angle of photos possible of the space. Building in public with the weird new tools.

45 million Gaussians. A sprawling medieval town, walkable in a browser tab. 🤯 Trogir, Croatia - running on @playcanvas' WebGPU 3DGS renderer - quick enough to keep a huge scene like this fluid as you explore. Sped up x2 and birds eye view at the end 🐦 [1/2]

Has anyone ever inspected a shoe on the web like this? #playcanvas

Finally you can inspect the number of splats in your SuperSplat scene! Currently it's only calculated for newly uploaded scenes, but we will backfill for the older ones.

It takes 42s to render the latest Blender splash art using Cycles on my machine. That's ~0.02FPS and clearly not suitable for real-time applications. But what if we wanted to? @SpenserFX recently described how they created a 3D Gaussian Splat (follow him!) from the Blender file and they kindly uploaded the splat to SuperSplat, which was then compressed into a 50MB .ssog file. Now anyone with a browser can inspect this masterpiece from any angle in real time (thanks to the powerful @playcanvas engine). Link in the comment below (download available).

Water shaders seem to be a thing. Here's a nice @playcanvas spectral one with caustics. marklundin.github.io/water/

I've updated the stochastic renderer in #SuperSplat editor to cull gaussians based on the previous frame's depth buffer. This scene from @varjodotcom has 85 MILLION gaussians and is running at nearly 100 FPS! (Watch the GPU graph bottom left - especially when trees block the view). No LOD or streaming. Just a massive scene with depth culling. Trying a similar approach for the forward renderer next!

Took a few hours to run these WebGPU benchmarks, so I'm sharing FYI: 🥊 PlayCanvas 🥊 Three.js 🥊 Babylon.js - @playcanvas: the winner in performance - @threejs: smallest + quickest to first frame - Babylon.js: still good tech but 🥲 Full report 👉 endel.github.io/webgpu-webgl…

I've added overdraw display to #SuperSplat editor. Runtime performance is heavily dependent on overdraw, so it's definitely a good idea to be aware of it in your scenes! And when comparing various trainers, overdraw is another quality metric to consider apart from PSNR! Right @lichtfeldstudio?

Free tool: SuperSplat Editor 3.0 is a big update The #opensource tool for editing 3D Gaussian Splats in a web browser has been rewritten on WebGPU, making it possible to visualize tens of millions of Gaussians smoothly cgchannel.com/2026/09/playca… #SuperSplat #3DGS @playcanvas

Same game, new lick of paint, plus a new sky, water, lighting, player model and HUD. Visual overhaul for my @playcanvas FPS is live. Browser, no download: ahoithere.com 👇

Apollo 11, simulated at full Earth-Moon scale in the browser. Real mission sequence, staging + docking, Google 3D Tiles for Earth, original NASA audio, and you can walk on the Moon. Built on arrival.space. and @playcanvas

I don't remember Quake 3 having such good graphics (built on @playcanvas)

3DGS動画の編集画面とレンダリング結果 コツコツとPlayCanvas版のSplatStrollAvatarを作っています。 SparkのRADフォーマット(http rangeリクエスト）と比較して、PlayCanvasのStreamed SOGはCDNとブラウザキャッシュの利用効率で明確にメリットあり。 3DGS表示品質も上で軽い。