Pusa is an advanced video diffusion model offering multi-task support, including Text-to-Video generation, Image-to-Video transformation, and frame interpolation. Its unique architecture enables frame-level noise control, improving video quality and efficiency.