Skip to content

Video Language Planning

Visual planning for complex long-horizon tasks

Visit official site

What Video Language Planning is

The algorithm is described as one that trains a vision-language model and a text-to-video model to achieve visual planning for complex long-horizon tasks.

What you can do with it

  • Train a vision-language model and a text-to-video model
  • Generate a detailed multimodal plan
  • Synthesise long-horizon video plans
  • Turn plans into real robot actions

Example uses

  • Stack objects in the centre of a table
  • Put fruit into the top drawer
  • Group blocks by colour

Pros & cons

✓ What we like

  • Generated video plans can become real robot actions
  • Plans span multi-object rearrangement to dual-arm manipulation

! What to watch out for

  • Input requires a long-horizon instruction and an image observation
  • Turning plans into actions needs a goal-conditioned policy

FAQ

What input does it take?

The material names a long-horizon task instruction and a current image observation.

What does it output?

The description states a detailed multimodal plan of video and language.

Last reviewed: 2026-11-15

More AI coding tools tools

View all →

How we review