Adobe introduces an experimental tool called MotionStream. Its goal is to solve two common problems in AI video generation: the lack of precise control and the slowness of the process. Instead of relying solely on text prompts, it allows users to manipulate objects and camera angles in real-time, using intuitive controls. This interaction aims to make motion creation more natural and eliminate the wait for full renders to adjust details.
Direct Manipulation Interface and Adaptive Generation 🎮
The technology is based on a user interface that translates cursor movements and slider adjustments into generation parameters in real-time. The system interprets these actions as direction and force vectors, integrating the changes into the frame flow without stopping the process. This allows for correcting trajectories, modifying element speed, or changing shots on the fly. The technical approach prioritizes low latency to maintain the illusion of direct control over the generated scene.
Goodbye to Novel-Length Prompts to Move a Cup ☕
With this, the era of writing prompts like a coffee cup that slides 5.3 centimeters to the right with a gentle roll, while the camera performs a 30-degree low-angle pan is over. Now we can push the darn cup with the mouse and see what happens. It's a notable advancement for those who thought animating a simple object required the skills of a playwright and the patience of a Buddhist monk.