Skip to content

WorldDreamer

Generate video by predicting masked visual tokens

Visit official site

What WorldDreamer is

The model is described as a video generation model that predicts masked visual tokens to understand and simulate world dynamics through unsupervised visual sequence modelling.

What you can do with it

  • Convert a still image into a moving video
  • Generate video from a text description
  • Inpaint, restyle or drive video from action input

Getting started

  • Visit the site and choose a generation type
  • Upload an image or enter a text description
  • Set parameters such as video style and length

Pros & cons

✓ What we like

  • Several generation types are listed in one model
  • Text and action inputs can be combined

! What to watch out for

  • The approach maps visual input to discrete tokens
  • Generation follows the selected video type

FAQ

How does it model the world?

The material describes predicting masked visual tokens from discrete token mappings.

What input types are named?

Images, text descriptions and action instructions are listed.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review