I wanted to know if Stable Video Diffusion can be used for collapsing buildings and destruction scenes. Taking a few keywords and numbers into account, SVD created these scenes overnight. No 3D models, smoke simulations, or manual animations were required to create this video, just AI.
Conclusion: As you can see, the scenes are mostly stationary. The process of a building collapse cannot yet be “simulated” because the system is unable to determine which parts of a structure would have to collapse. But I can imagine that future AI models will be able to even estimate cause and effect.
Credits:
Music: UNIVERSFIELD - Atmosphere for Documentaries (CC BY)
Software:
ComfyUI
Stable Diffusion XL Turbo
Stable Video Diffusion