4DAnyone turns one casual video into 48 consistent camera views for 4D Gaussian Splatting
jonstephens85 · x · 2026-09-05
jonstephens85 tested 4DAnyone, generating 48 consistent camera feeds of himself from a single casual video capture, targeting 4D Gaussian Splatting from one camera. He visualized progress with Rerun and says he hacked the visualization together with ChatGPT's new Astra model in VS Code in under 10 minutes. Pablo Vela Gómez's 6-view 4DAnyone generator is also live on Hugging Face.
More from Multimodal
- Take a photo of the real world and paste it straight into Blender for 3D — sidahuj · 2026-09-05
- Astra researched Apple Park's trees and modeled each one, showing striking 3D fidelity — jxnlco · 2026-09-05
- GPT-6 Astra Reproduces Images Pixel-by-Pixel, Mocked as 'Reinventing JPGs' — alejandroll10 · 2026-09-05
- Every Clip in This AI-Hosted Late Night Show Was Generated with Veo — presentofai · 2026-09-05
- Open-Source LoRA Adds VR180 Stereoscopic SBS Output to MiniMax H3 — Ok_Tale7582 · 2026-09-05
- ECCV 2026 launches VisPer workshop on persuasive visual reasoning and generation in multimodal AI — simi_97k · 2026-09-05