Skip to content

CausVid

Generate video frames as a causal stream

Visit official site

What CausVid is

The model is described as a video generation model that adapts a pretrained bidirectional diffusion transformer into a causal transformer for instant frame generation.

What you can do with it

  • Stream high-quality video at 9.4FPS on one GPU
  • Distil a 50-step diffusion model into 4 steps
  • Synthesise long videos from short training clips

Application scenarios

  • Generate a clip of a snowman melting
  • Turn a text prompt about a paper plane into a swan into a five-second clip
  • Animate a retro-futuristic robot from a still image

Pros & cons

✓ What we like

  • Streaming generation at 9.4FPS on one GPU
  • Long video synthesis from short training clips

! What to watch out for

  • Built by adapting a pretrained bidirectional diffusion model
  • Distillation depends on a bidirectional teacher model

FAQ

What frame rate is stated?

The material states 9.4FPS on a single GPU.

What inputs does it take?

Text prompts and zero-shot image-to-video input.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review