https://github.com/bring-shrubbery/ml-sharp-web #AI #GaussianSplats #GitHub #BuzzwordBingo #TechNews #HackerNews #ngated
*whoa*
ml-sharp (https://github.com/apple/ml-sharp) works especially well with messy images, where the gaps are not so noticeable.
Apples ml-sharp now with a custom gaussian splat renderer in a Tiefling branch (https://github.com/combatwombat/tiefling/tree/test/sharp-splat). Can't be published on the web though, too fat and wrong license.
But it works well with all kinds of images. One step closer to a holodeck :D
Apples ml-sharp is pretty impressive (https://github.com/apple/ml-sharp). It creates Gaussian splats from one image, in a few seconds.
First part of the video, the old depthmap + parallax rendering in https://tiefling.app. Edges are wonky, camera movement is limited.
Second part of the video, exploring the generated splat in https://superspl.at/editor and flying through the image o_o
At Magnopus, we do a lot of work with Gaussian Splats, but as a backend/infrastructure engineer, I never really understood how they worked. Until now!
https://youtu.be/X8yRlA7jqEQ?si=PP-MAL0GaE3EtWXT
#gaussiansplatting #gaussiansplat #gaussiansplats #3d #3drendering


I'm not sure I agree with Wren about the use cases at the end, but this is still an extremely interesting technology for film/video creators.
Also, can't help thinking that this video coming out the same week Sora shut down is a stark illustration of something we should know by know: Shortcut products are a dead end - where technology really shines is well-thought-out tools for well-defined applications.

AA (@measure_plan)
가우시안 스플랫(gaussian splats)을 탐색하는 인터페이스 개인 프로젝트 소개: World Labs의 'marble'로 9개의 세계를 생성하고, 각 세계를 기하학적·추상적·광활함·친밀성 등 의미론적 특성 기반 그래프로 시각화했으며, MediaPipe 컴퓨터 비전을 이용해 얼굴 추적 카메라 컨트롤을 추가한 구현입니다.
https://x.com/measure_plan/status/2029229021070774433
#mediapipe #computervision #visualization #gaussiansplats #3d

i made an interface for exploring gaussian splats - 9 worlds created with World Labs marble - visualized in a graph based on semantic understanding of each world (geometric vs abstract, vast vs intimate) - added mediapipe computer vision for face-tracked camera controls and
XGRIDS (@XGRIDS_OFFICIAL)
EvoRealitycraft가 XGRIDS로 캡처한 Gaussian Splats를 Arrival.Space에서 워커블하고 인터랙티브한 환경으로 변환하는 데모를 공유했습니다. 실세계 캡처와 XR 실험을 결합해 장소와 시간, 스토리텔링을 혼합하는 창작 워크플로우를 지원하는 사례입니다.
When real-world capture meets #XR experimentation. @EvoRealitycraft demonstrates how Gaussian Splats captured with #XGRIDS in https://t.co/9An73e8kOt can become walkable, interactive environments—supporting creative workflows that blend place, time, and storytelling.