Skip to content

Show-1

Combine pixel and latent diffusion for text-to-video

Visit official site

What Show-1 is

The model is described as an efficient text-to-video generation model that combines pixel-level and latent-level diffusion models. It first produces a low-resolution initial video with a pixel-level model, then upsamples it to high resolution with a latent model.

What you can do with it

  • Generate videos that closely follow the input text
  • Produce high-quality video output
  • Work with low compute resource requirements

Example uses

  • Enter a panda beside the waterfall holding a sign
  • Enter a girl dancing in the room to generate her dance video
  • Enter a cat playing with yarn to generate the video

Model notes

  • A pixel-level model creates the low-resolution initial video
  • A latent-level model upsamples the result to high resolution
  • Text relevance is stated as closer than a pure latent model
  • Compute cost is stated as lower than a pure pixel model

Pros & cons

✓ What we like

  • Closer text alignment is stated compared with a pure latent model
  • Lower compute cost is stated compared with a pure pixel model

! What to watch out for

  • Generation runs through two stages: a pixel model and a latent model
  • The named scope stays with text-to-video and conditional video generation

FAQ

How does it generate video?

The material states a pixel-level model makes a low-resolution initial video and a latent model upsamples it to high resolution.

What prompts are listed as examples?

The examples cover a panda beside a waterfall holding a sign, a girl dancing in a room, and a cat playing with yarn.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review