Developers can control generated video environments using WASD inputs.
Developers can control generated video environments using WASD inputs.
@bilawalsidhu Thanks man!!
Lingbot-World-2, the new real-time world model just arrived on @reactorworld. Here, you control the scene and the world while is still forming. You move the camera, and the model generates a responsive world. A normal video model makes a clip, but this system keeps rendering while you move through it. Reactor is the infrastructure layer for real-time World Models, it brings us a real-time navigable world model with API access. So developers can now build products where the world changes as the user acts. World Models change the job of video from playback to live generation, where pixels are created as the user acts and speaks. A few lines of the Reactor SDK and you're streaming pixels from a World Model into your product in real time. Huge for games, creative tools, simulation, robotics, storytelling, and many other categories. With Lingbot-World-2, it takes a reference image and text prompt, then generates an interactive video stream that can be steered with WASD movement, arrow-key looking, prompt changes, and native camera-pose controls.
This world model is impressing many, including me. Longer video length means that the companies are figuring out how to make the world models more efficient. Now, in 18 months, what do you think that leads to? How about two video streams, one going to each eye in a pair of glasses? And let's see the quality go way up. As impressive as this is, it's still very early in world model history.
Okay this live demo is zippy af -- give it a try! I'm gonna have to vibe code some dopeness with all these world model APIs dropping left and right. https://twitter.com/reactorworld/status/2074935025771143626
Try it here https://www.reactor.inc/lingbot-world-2
