Alexandre Devaux

@gradeeterna 🔥🔥🔥

Testing my DJI Osmo 360 rig for gaussian splatting. Mason’s Avenue, London. 14 million splats total. Trained with LichtFeld Studio and gsplat, edited in Houdini GSOPs, camera animation in Unity and rendered with Deckard Render. #gaussiansplatting #3DGS

@junyi42 🔥🔥🔥

𝗢𝗻𝗲 𝗺𝗲𝗺𝗼𝗿𝘆 𝗰𝗮𝗻’𝘁 𝗿𝘂𝗹𝗲 𝘁𝗵𝗲𝗺 𝗮𝗹𝗹. We present 𝗟𝗼𝗚𝗲𝗥, a new 𝗵𝘆𝗯𝗿𝗶𝗱 𝗺𝗲𝗺𝗼𝗿𝘆 architecture for long-context geometric reconstruction. LoGeR enables stable reconstruction over up to 𝟭𝟬𝗸 𝗳𝗿𝗮𝗺𝗲𝘀 / 𝗸𝗶𝗹𝗼𝗺𝗲𝘁𝗲𝗿 𝘀𝗰𝗮𝗹𝗲, with 𝗹𝗶𝗻𝗲𝗮𝗿-𝘁𝗶𝗺𝗲 𝘀𝗰𝗮𝗹𝗶𝗻𝗴 in sequence length, 𝗳𝘂𝗹𝗹𝘆 𝗳𝗲𝗲𝗱𝗳𝗼𝗿𝘄𝗮𝗿𝗱 inference, and 𝗻𝗼 𝗽𝗼𝘀𝘁-𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻. Yet it matches or surpasses strong optimization-based pipelines. (1/5) @GoogleDeepMind @Berkeley_AI

@Haian_Jin Brilliant work! Looking forward to trying the code 🔥

Spatial reconstruction is a long-context problem: real scenes come with hundreds of images. But O(N²) transformer-based models don’t scale efficiently. Introducing: 🤐ZipMap (CVPR ’26): Linear-Time, Stateful 3D Reconstruction via Test-Time Training (TTT). ZipMap “zips” a large image collection into an implicit TTT scene state in a single linear-time operation. The state will then be decoded into spatial outputs, and can be queried efficiently for novel-view geometry and appearance (~100 FPS) ZipMap is not only much faster (>20× faster than VGGT), but also matches or surpasses the accuracy of all SOTA models.

@mrdoob @deno_land A camera near a bit too far ? 😉 That's nice!

@deno_land Kind of want to keep this bug 🤓

Okay Claude, can you help me port Quake to Three.js? ... One hour later

@Jaenam97 Superb!

Shader-magic 🪄 Rendering flowers from a dodecahedron wireframe 🔮 #glsl

@onirenaud Exciting work! 🔥 Following closely, When will be able to test it?

Got Gaussian Splatting to 120fps for ~200k splats with: - Morton order for tight chunk bounds - Hierarchical chunk culling with indirect dispatch - 16bit Radix sort - Render Bundle to cut CPU overhead - Packed buffers Now trying 120fps with ~500k, bottleneck still overdraw.

@YuanLiu41955461 Nice!! Thanks for sharing the code 😀

Excited to share our recent work, UniSH, which unifies dynamic 3D scene reconstruction and SMPL estimation within a single framework. (Left-top is input video). Code has been released! github.com/murphylmf/UniSH Project page: murphylmf.github.io/UniSH/ Paper: arxiv.org/abs/2601.01222

@skalskip92 Congrats! Segmentation is pretty hot to me, just tested on your hugging space. Can I use it for images used later in broadcast without complexe rights/licence? 😀

RF-DETR paper is finally on arXiv - real time detection with DINOv2 backbone - runs neural architecture search (NAS) over about 6000 architecture variants - uses weight sharing across all configs - first real-time segmentation DETR to break past top YOLO results ↓ more

@romvsuals Some surprising relaxing good feeling about this flat. Well done!🔥

Wip 💻

@romvsuals Some surprising relaxing good feeling about this flat. Well done!🔥

@YuanLiu41955461 🔥🔥🔥 Looks impressive! Any Hugging face demo coming ? 😀

Share our recent work - TrackingWorld: World-centric Monocular 3D Tracking of Almost All Pixels [NeurIPS'25] An optimization-based 3D tracking of all pixels. Code: github.com/IGL-HKUST/Trac… Webpage: igl-hkust.github.io/TrackingWorld.… Welcome to try it! A feedforward version is on the way!

@skandakoppula So cool! 🔥🔥🔥

Thrilled to release 🎯 D4RT (Dynamic 4D Reconstruction and Tracking)! 🌟 State-of-the-art results on 4D reconstruction & tracking benchmarks 🚀 Up to 300x faster tracking and 100x faster pose estimation than prior works 📍 A simple, unified interface for tracking, depth, and pose using point-wise decoding 🔗 Learn more about D4RT: d4rt-paper.github.io

@pushmatrix @Shopify So cool, my dream came true! 😀 x.com/AlexandreDevau…

Can you run WebGL shaders in realtime on Sphere? Turns out you can. @Shopify just did it for visualizing realtime sales in a whole new way.

@pablovelagomez1 @rerundotio @Gradio Well done :)

Finished building out the @rerundotio and @Gradio app for SAM3D-Body, and I think it came out really clean! Under the hood, it's using three models 1. sam3d for exemplar segmentation based on the "person" text prompt. 2. sam3d-body for generating the 2d keypoints, 3d keypoints and mesh 3. mogev2 for intrinsic/fov estimation Really happy with how it came out. I'll probably start working on videos and multiview captures next!

@Machinedelusion Congrats :)

Eh! Thanks so much for the Corridor Crew for shouting me out and my tutorial on their latest video! You can watch their video here youtube.com/watch?v=ct_7FU…

@bingyikang Even with a Hugging face demo. Thx!! 🔥🔥🔥

After a year of team work, we're thrilled to introduce Depth Anything 3 (DA3)! 🚀 Aiming for human-like spatial perception, DA3 extends monocular depth estimation to any-view scenarios, including single images, multi-view images, and video. In pursuit of minimal modeling, DA3 reveals two key insights: 💎 A plain transformer (e.g., vanilla DINO) is enough. No specialized architecture. ✨ A single depth-ray representation is enough. No complex 3D tasks. Three series of models have been released: the main DA3 series, a monocular metric estimation series, and a monocular depth estimation series. The core team members, aside from me: @HaotongLin, Sili Chen, Jun Hao Liew, @donydchen. 👇(1/n) #DepthAnything3

@xxunhuang Very impressive, great work! 🔥

It feels totally different when video models actually become real-time interactive. Drag the mouse, and the cup instantly moves with it, and the water follows. You’re not just watching a video anymore, you’re playing with it.

We present MotionStream — real-time, long-duration video generation that you can interactively control just by dragging your mouse. All videos here are raw, real-time screen captures without any post-processing. Model runs on a single H100 at 29 FPS and 0.4s latency.

@kwangmoo_yi Great work! 🔥 CVPR 2026 submission?

Zhou et al., "PAGE-4D: Disentangled Pose and Geometry Estimation for 4D Perception" VGGT extended to dynamic scenes with a dynamic mask predictor.

@onirenaud So good!

This is what happens when you combine two features of Three.js Blocks. The SPH simulation is rendered as an InstancedMesh and parented to a MeshTransmission. By simply rotating the parent, the magic unfolds. Can’t wait to see some snow globe experiences this Christmas! ❄️✨

@0xca0a It still runs super smooth. Insane 🔥

trying out the upcoming ssgi effect (paired with ssr). i guess all i ever wanted out of three is reflections and global illumination. come next three we'll have it. 🫡 live demo link (+ code) below ↓