Henry Ruhs retweeted
写真からオリジナルのMetaHumanの顔を作れる「KeenFace」に注目
開発者(KeenTools氏)が新しいWebツールを使った制作デモを公開しました。参照画像をもとに人物の顔を再現し、UE5で使うキャラクター制作につなげる。実在する人物をモデルにしたキャラづくりの選択肢として、活用が気になるツールです
Henry Ruhs retweeted
Here's a free City Generator that can export to GLB, OBJ, or FBX for your MCP Blender projects.
Even if you're NOT going to use it for 3d scenes, it's just fun to play around with for 5 minutes!
Henry Ruhs retweeted
I am training 3D Gaussian splats in the browser in real-time, no uploads or servers!
Wrote the whole pipeline in rust (surprise!) using Wasm / WebGPU. Runs ~50% faster on a native / Vulkan backend too.
Henry Ruhs retweeted
the cybersecurity industry is completely disrupted.. 💀
someone really open-sourced a fully autonomous AI Red Team.
it’s called PentAGI. multiple AI agents that talk to each other to hack a target. zero human input.
100% open-source.
Henry Ruhs retweeted
I MADE IT!
I beat Adobe and shipped a better relight feature than them. (theirs is not even released yet).
A multibillion-dollar company with thousands of engineers just got beat by one person.
Crazy thing is that ImageFlow.dev runs in your browser locally for Free 🤯
Let's kill Adobe one feature at a time.
Stay tuned, I'm working on a desktop version of ImageFlow and a new LightRoom alternative that i'm so excited about!
One year of development has led to this moment: FaceFusion v4 beta is available now, and it is by far the biggest step in the history of the project. To celebrate the release, we are kicking off the **FaceFusion v4 Challenge** starting today.
Henry Ruhs retweeted
⭐️I am very happy to present to you the brand Release Trailer of HAELE 3D - Portrait Studio Pro⭐️
- Tons of blendable facial expressions
- Planar heads, skulls, facial muscles
- Fully editable realtime lighting
📅Release Date: 09/23/2026
Would you use it in your practice?
Henry Ruhs retweeted
Face Anything: 4D Face Reconstruction from Any Image Sequence
ECCV 2026
github.com/kocasariumut/Face…
Face Anything is a unified feed-forward model for high-fidelity 4D face reconstruction and dense tracking from arbitrary image sequences. The key idea is canonical facial point prediction, a representation that assigns each pixel a normalized facial coordinate in a shared canonical space. This formulation transforms dense tracking and dynamic reconstruction into a single canonical reconstruction problem, producing temporally consistent geometry and reliable correspondences.
Henry Ruhs retweeted
kimodo.cpp is up on GitHub! It allows you to animate a skeleton by describing what you want it to do.
Henry Ruhs retweeted
AntResearch just dropped 4DAnyone on Hugging Face!
Convert a video of any person into a 4D gaussian splat for that person
▶️ huggingface.co/spaces/huggin…
Henry Ruhs retweeted
LUNA: Learning Universal 3D Human Animation Beyond Skinning
ECCV 2026
penghtyx.github.io/LUNA/
Creating photorealistic, animatable 3D human avatars from monocular images still largely depends on Linear Blend Skinning (LBS) and parametric body models which constrain expressivity and often introduce artifacts due to imperfect fitting. We propose LUNA, an LBS-free universal neural animation model that directly maps diverse 2D controls like images, keypoints, sketch and unseen characters into 3D Gaussian deformations, bypassing explicit body fitting. At its core, a transformer-based motion regressor disentangles global rigid motion from fine-grained local dynamics to capture both coherent movement and subtle non-rigid effects. To resolve the inherent ambiguity of 2D-to-3D lifting while scaling beyond fitted datasets, we introduce hybrid supervision that distills soft structural priors from an LBS teacher and a loss that supports training on both limited fitted data and large in-the-wild unlabeled videos. Extensive experiments show LUNA achieves competitive visual fidelity compared to LBS-based approaches, while delivering realistic human motion and zero-shot cross-identity generalization across diverse driving modalities. To the best of our knowledge, LUNA is the first end-to-end 3D animatable model that supports implicit 2D driving.
Henry Ruhs retweeted
I’ve been experimenting with creating 3DGS portraits (for now, selfies) from a single photograph.
The reconstruction is far from clean or geometrically accurate, but I really like this noisy radiance field aesthetic: the unstable geometry, the translucent fragments, the floaters suspended around the subject. I liked those artifacts in the early days of NeRFs, and I still do. My next step will be to work more deliberately with those fragments and floaters, rather than trying to eliminate them.
The image is turned into a synthetic 360° video, converted into an image sequence, aligned in RealityScan, and finally trained as a Gaussian splat with Litchfield.
#3DGS #AI #selfie #fail
Henry Ruhs retweeted
duardo Alvarado(@roboonaut)氏によるGoogleの3D顔モデル「GNM」とMediaPipeを組み合わせたフェイストラッキングの試作プロジェクト。
GNMヘッドで記録したランドマークと実画像の検出結果を比較し個人差をフィット。
まだ粗いもののリアルな顔追従を実現。
Henry Ruhs retweeted
Topologically Consistent Multi-view 3D Head Reconstruction via Coarse-Guided Layered Surface Sampling
Google
SIGGRAPH 2026
syntec-research.github.io/SH…
TL;DR: From calibrated multi-view images, SHELLS reconstructs 18k-vertex 3D heads in 0.08 seconds. It aggregates DinoV2 features via projective surface-aware feature sampling, allowing a transformer to predict dense semantic meshes 3.5x faster with 88% less GPU memory than state-of-the-art methods. The result is an efficient, globally consistent multi-view reconstruction framework.
Henry Ruhs retweeted
THIS IS CRAZY 🤯Bifrost just shipped a proper real2sim pipeline that actually works at production speed!
You send a single video of a real industrial site or lab. Their system reconstructs a fully navigable, photorealistic 3D environment in ~30 minutes instead of the usual 12+ hours of manual modeling.
Technical stack under the hood:
- 3D Gaussian Splatting (and derivatives like World Labs Marble) turns the video frames into a dense, view-dependent radiance field. This gives real-time rendering of the exact geometry, materials, and lighting of the captured space without traditional mesh reconstruction bottlenecks.
- Articulated asset generation (building on models like TRELLIS for static objects and Articraft-style systems for movable parts) automatically extracts and converts objects into simulation-ready rigid-body or articulated assets with correct collision geometry, joints, and physical properties.
- GPU orchestration + domain randomization layers on top so the environment can be relit (GaRe-style), cluttered, and sensor-simulated across RGB, depth, segmentation, and 6-DoF poses.
The result is a high-fidelity digital twin that robots can interact with immediately. You drop your policy into the reconstructed scene, run thousands of parallel rollouts on Manifold, and surface failure modes at compute speed rather than waiting for real-world iterations.
This closes a big part of the classic sim-to-real gap: visual fidelity no longer looks like a 2001 PS1 game, and the physics/assets are grounded in the actual site instead of generic asset libraries.
The DeLorean garage demo is just the public teaser: the same pipeline is already being used for factory cells, warehouses, and manipulation benchmarks.
Henry Ruhs retweeted
Crystal is now the largest image-upscaler API in the world
I've increased the output limit to 1.5 gigapixels, which is 38K 😂

















