world models
Every SLOP TV News story about world models.
Google's Genie 4 makes photoreal worlds you can ride and hear
DeepMind's world model generates the sound with the scene, takes typed instructions mid-simulation and holds a world together for two minutes, double Genie 3. Trusted testers are already using it to prototype shots.
EdenLive renders 720p video at 24.7 FPS on one RTX 5090
The open-weight renderer's code and main weights are Apache 2.0, but the streaming service needs an 80 GB NVIDIA GPU and CUDA 13.
Matrix-Game 4.0 renders game worlds at 60 FPS on one RTX 5090
Skywork AI's open-weights release splits an executable game engine from a causal video model and ships three playable scenes under Apache-2.0.
SolarRenderer turns game-engine block-outs into video in 4 steps
The team froze a 33B MiniMax-H3 video model, added two small LoRAs, and cut sampling from 50 steps to four, with a streaming variant at 768p.
NVIDIA Cosmos3-H-Dreams generates surgical video at 43.2 fps
The model learned from CMR Surgical Versius gallbladder-surgery recordings and runs on one RTX PRO 6000, but NVIDIA calls it a research simulator.
Viggle built a 3D village in one second, and I am sick with envy
A Toronto company taught a machine to put a boy in a village and walk him across it, physics built in, for less than a cent a scene. I have been typing for fifty years, and I cannot do it in six weeks.
Amazon in talks to buy AI video startup Decart for $7B
The Wall Street Journal reports a roughly 75 percent premium on Decart's May valuation, weeks after Anthropic walked away from a $6 billion deal.
fal Worlds streams walkable AI worlds for $4.50 an hour
Lingbot World 2 and Abot World 0 both stream over WebRTC, and the faster one is research only while the Apache-licensed one is cleared for commercial work.
ByteDance's world model re-shoots a movie scene, and Krea 3 lands
Every segment of Theoretically Media's latest episode, checked against the sources it links: a nameless ByteDance world model turning one drawn camera path into a dozen angles inside a Stephen Chow film, Pico's delayed Vision Pro rival, Krea 3's eight-second edits, the report that Sora comes back through a filmmaker startup, an OpenAI engineer charged over a hidden recorder, and two free tools.
Helix World generates video and spatial audio in one pass
Noiz AI and HKUST built the world model on the open-source LTX-2.3 model and posted a preview checkpoint, but local runs want an 80 GB GPU.
MBZUAI's WorldGuide makes AI video plan its own next steps
The UAE university's model plans each step from the clip it just generated, scoring 33.33 percent task success against 29.90 percent for MiniMax-H3.
Seedance 2.5 tops an AI video physics test at 57.76 out of 100
A new benchmark ran 40 controlled physics tasks across eight video models and 1,280 clips, and only one model in the set is closed.
WorldPlay2 turns one image into a playable browser world
The weights landed on October 6 under a non-commercial license, and a hosted version went live on Reactor the next day.
Runway's Praxis-1 turns video pretraining into robot control
The New York company says early partners Noble Machines, Standard Bots and Ultra are running Praxis-1 on their own hardware, with open weights planned for the coming months.
Project Continuum guide: Runway's AI operating system explained
Runway Labs wants your computer to render its own interface instead of shipping apps. Its own build calls itself a real-time video operating system at v0.0.1-alpha. Here is what it does, what it runs on, and what is still broken.
World-model startup General Intuition raises $220M at $6.2B
A year after spinning out of game-clip platform Medal, the lab says its MIRA model renders training footage at 20 frames per second across four players.
AMD buys Fei-Fei Li's World Labs for $8.2 billion
The chipmaker gets the spatial-intelligence lab behind Marble, and founder Fei-Fei Li joins AMD as executive vice president and chief scientist.