Description

State-of-the-art generative video models. We dissect how 2D spatial diffusion is extended into 3D spacetime via temporal attention blocks and motion vectors.

Key Topics

  • Temporal Consistency: Overcoming frame flicker and semantic drift using latent temporal self-attention.
  • Camera & Motion LoRAs: Pan, zoom, tilt, and trajectory control in generative video.
  • Generative Cinematic Language: Editing, camera staging, and conceptual filmmaking with AI tools.

References & Wiki Links