A humanoid robot policy trained solely on synthetic data generated by a world model. Research Scientist Joel Jang presents NVIDIA’s DreamGen pipeline: ⦿ Post-train the world model Cosmos-Predict2 with a small set of real teleoperation demos. ⦿ Prompt the world model to https://x.com/TheHumanoidHub/status/1945569893048619201
Eric Jang explains that 1X’s world model is a video generation model at its core, but had to be retrained from scratch to make it action-controllable. He also notes that simulated people in the world model might eventually pass the Turing test. https://x.com/TheHumanoidHub/status/1944975679692460288
Eric Jang explains why 1X is building a world model and what it can do that rigid-body simulators can’t. https://x.com/TheHumanoidHub/status/1943555624501121467
This is a real job now. Build the waifu of your dreams at @xAI. https://x.com/ebbyamir/status/1945247680176799944
Industry video game actors pass agreement with studios for AI security | Reuters https://www.reuters.com/business/media-telecom/industry-video-game-actors-pass-agreement-with-studios-ai-security-2025-07-10/
Her hand is clipping through her thigh, and her character card is full of typos. this shows absolutely unacceptably low standards in waifu engineering. Mihoyo would never allow this slop. Elon needs gooners on board. The sort of people who collect lewd plastic figurines. https://x.com/teortaxesTex/status/1945737831697064446
SpatialTrackerV2: unified, end-to-end 3D point tracking model which simultaneously estimates Camera Motion, Consistent Geometry and Pixel-wise 3D Trajectories. https://x.com/bilawalsidhu/status/1945154158782505345
AI image generation 3D “”game engines”” seem to have reached some critical ability threshold recently (not there yet of course, but much improved over a couple months ago). Here is another one, Odyssey, that lets you wander around virtual worlds where each frame is AI generated. https://x.com/emollick/status/1943354179260682652
Combining the explicit control of 3D software with the creativity of generative AI models is a promising yet underrated workflow. Build your 3D scenes procedurally by describing them in natural language, then take them all the way with your image & video models of choice. Tools https://x.com/bilawalsidhu/status/1945619824358690993
DreamCube: 3D Panorama Generation via Multi-plane Synchronization https://yukun-huang.github.io/DreamCube/
DreamCube: 3D Panorama Generation via Multi-plane Synchronization”” TL;DR: Multi-plane Synchronization to adapt 2D diffusion models; multi-plane panoramic representations (cubemaps); facilitates tasks including RGBD panorama generation and depth estimation, 3D scene generation. https://x.com/Almorgand/status/1943693771599745368
MILo: Mesh-In-the-Loop Gaussian Splatting for Detailed and Efficient Surface Reconstruction – YouTube https://www.youtube.com/watch?v=_SGNhhNz0fE
MILo: Mesh-In-the-Loop Gaussian Splatting for Detailed and Efficient Surface Reconstruction”” TL;DR: differentiably extract a mesh including both vertex locations and connectivity only from Gaussian parameters. Gradient flow from the mesh to GS https://x.com/Almorgand/status/1943413424328704411
We can just generate cities using open source 3d software and cheap plugins https://x.com/bilawalsidhu/status/1944460975824048419
SoulDance https://xjli360.github.io/SoulDance/
StreamME Project Page https://songluchuan.github.io/StreamME/
MILo: Mesh-In-the-Loop Gaussian Splatting for Detailed and Efficient Surface Reconstruction https://arxiv.org/pdf/2506.24096
SceneScript treats 3D reconstruction as a language problem rather than a geometry one. The model watches a video of a room and just learns to write a script for it. It autoregressively spits out text commands like make_wall(…) or make_bbox(…) that define the scene. https://x.com/bilawalsidhu/status/1944760878831923522
