Zarbuz (@Zarbuz) header
Zarbuz (@Zarbuz) profile picture

Software Engineer, Voxel Artist and Admin of @MVC_discord Secretly working on solution for collection based on voxels

Paris, France
Joined January 2013
Photo by Zarbuz (@Zarbuz)Photo by Zarbuz (@Zarbuz)Photo by Zarbuz (@Zarbuz)Photo by Zarbuz (@Zarbuz)Photo by Zarbuz (@Zarbuz)Photo by Zarbuz (@Zarbuz)Photo by Zarbuz (@Zarbuz)Photo by Zarbuz (@Zarbuz)Photo by Zarbuz (@Zarbuz)Photo by Zarbuz (@Zarbuz)
In case you missed it, the WebGPU meetup is happening in Paris Nov 5th. CFP is still open until next Wednesday! Survey will drive prioritization for the standardization group and your input will be extremely valuable (bindless is already in progress ^^) Links in the thread 🧵👇
🚀 The WebGPU Working Group is meeting in Paris this November, and they need your real-world insights to shape the future of the standard! Tell us what works well and what needs improvement 👇 📊 Take the survey: forms.gle/H57wGhsTFjQbeBTv7 [1/3]
4
11
3,109
I asked Opus 4.6 to 5.5 to generate a realistic 3d model of a medieval castle. It's scary how much it has improved in the past 7 months.
63
100
12
2,473
214,609
Zarbuz retweeted
I reverse engineered and reimplemented DLSS 5's neural rendering in Vulkan, from the architecture down to the last rounding. same input, same output as Nvidia's DLSS5, 7.8ms per 1080P on RTX 4070 super code and docs : github.com/maanHimself/OpenD…
37
137
16
1,596
90,447
Zarbuz retweeted
DLSS 5 running in the browser with #webgpu And yes it works on MacOS too. Try the the live demo here dlss5-webgpu.dlss5-webgpu-st… You can also try it with your own models #WebDev #AI #threejs
34
54
5
498
38,880
Splat.js has arrived. We just open-sourced a 3D Gaussian Splatting trainer that runs entirely in your browser. Platform-agnostic, powered by WebGPU. No server. No CUDA. No COLMAP. No upload. Photos or 360 panos in → camera poses solved → 3D Gaussian Splat trained against your photos at sub-pixel accuracy → standard .ply out → instant publishing to @arrival_space_ Everything happens locally. In one browser tab. Free. Open source. MIT licensed. Try it. Break it. Build on it. Link + GitHub repo in the comments. Thanks for the article @RadianceFields
6
45
3
210
10,073
🚀 The WebGPU Working Group is meeting in Paris this November, and they need your real-world insights to shape the future of the standard! Tell us what works well and what needs improvement 👇 📊 Take the survey: forms.gle/H57wGhsTFjQbeBTv7 [1/3]
2
1
2
10
18,614
Zarbuz retweeted
This is Fast Enter Play Mode - a setting that makes it significantly faster to enter Play mode, especially after small changes that used to trigger the full domain reload. On the path to CoreCLR, we are making this mode the default for new projects in Unity 6.6 and later.
6
21
6
312
23,886
Zarbuz retweeted
Introducing Unity 7 - built for a faster-moving, more open industry where creators, cross-functional teams, and coding agents can seamlessly collaborate across every stage of the development cycle. Featuring: - Create faster: Modernized CoreCLR foundation with a near-instant Play Mode and up to 90% faster shader builds. - Collaborate in better ways: New CLI, public API, and free MCP connect teams and coding agents directly to Unity. - Ship sharper graphics: Richer, realistic lighting via Surface Cache GI across all Unity supported devices. - Grow and monetize smarter: Built-in Unity Vector tools, native webshops, and direct-to-consumer IAP to scale efficiently. - No breaking changes: A direct continuation of Unity 6 architecture, offering an upgrade path with zero rebuilding. Unity 7 will enter early Beta testing in December, with the Full Release scheduled for Q1 2027.
61
395
108
2,126
621,357
Time to do some interesting things 🧐 #portalcam @XGRIDS_OFFICIAL
1
106
Zarbuz retweeted
3D scene reconstruction works great until the camera never sees part of the scene. ArtiFixer from NVIDIA Research is an open autoregressive model that fills in the missing geometry that other methods leave blank. #SIGGRAPH2026 paper, code + demo: nvda.ws/4oILqNd
32
169
21
1,312
155,880
Push constants? Root constants? In WebGPU, they’re called Immediates. ⚡️ Pass small, fast-changing data straight to your shaders. Zero buffer creation. Zero bind group overhead. Pure performance. 🏎️💨 Check out developer.chrome.com/blog/ne… to learn more
2
13
704
It's actually le gros chaton
418
793
302
9,236
1,858,601
🏆 Just launched FaceOff WC26 — a football guessing game for the World Cup. Guess a mystery player in 6 attempts. Clues revealed randomly after each wrong guess. Daily Challenge · Marathon · Quiz Team WC kicks off June 11 👀 👉 faceoff-wc26.expo.app/downlo… #WorldCup2026 #FIFA
1
2
3
419
SPZ 4 is here. 🚀 3–5x faster encoding. 2x faster loading. 20x faster in the browser. Tens of millions of points, no sweat. The "JPG for 3D Gaussian splats" just leveled up. Big thanks to everyone who made this release happen, especially the team at @Adobe. Read more: hubs.ly/Q04fvnx90 #NianticSpatial #Scaniverse #SPZ #GaussianSplat #Reconstruction #PhysicalAI #GeospatialAI #AI #Adobe
10
123
10
714
63,988
🚨 SplatTransform 2.0 is here!🚨 Open source CLI tool and library for 3DGS processing NEW: Fully automated, lightning fast creation of high quality collision for your splats. Outputs: 🧊 Voxel format 🔼 Mesh format See both formats in action in the video. Links 👇
5
73
2
488
36,381
I keep saying this but it needs to be said again, people already have a Switch 2 in their pockets, we just need AAA companies to realize that and start porting games for our phones........ Perhaps it would help to know Apple ships roughly 230 Millions iPhones per year, more than all consoles combined for the entire 7 years generation is done by a single phone company.
Ray tracing on iPhone 17 Pro is here thanks to @ControlRemedy Can’t believe it works well on a phone
1
2
217
📢GaussianGPT: autoregressive 3D Gaussian scene generation. We introduce a GPT-style model that directly generates 3D Gaussian scenes, token by token, in a series of small, discrete decision steps. Generation, completion, and large-scale outpainting in a single pipeline. Unlike diffusion-based approaches, GaussianGPT explicitly models the scene distribution at every step, allowing for quite flexible scene synthesis. 🌐 nicolasvonluetzow.github.io/… ▶️ youtu.be/zVnMHkFzHDg Great work by @nicolasvluetzow, @barbara_roessle, @katha_schmid
35
287
16
2,317
152,813