Motion Prompting ; Breakthrough in Video Generation from Google DeepMind ✨📹

✨ A breakthrough in video generation: Motion Prompting! 📹 from Google DeepMind

This approach leverages spatio-temporal motion trajectories as conditioning signals, giving users flexible control over video content like object and camera movements. 🎮🎥

Here’s how it works:

• ControlNet adapter trained on a pre-trained video diffusion model powers this flexibility.
• Motion prompt expansion translates high-level user requests into detailed motion trajectories, making complex tasks feel effortless.

The results?

• Diverse applications with satisfactory outcomes
• Emergent behaviors like realistic physics simulations 🌌⚙️

Paper
Github

#AI #VideoGeneration #MotionPrompting #Innovation #ControlNet #VideoDiffusion #CreativeAI #Google #Deepmind #GenerativeAI #GenAI #LLM #LMMM