Skip to content

SCENIC model

Navigate 3D scenes with text and user trajectories

Visit official site

What SCENIC model is

The model is described as a text-conditioned scene interaction model that navigates 3D scenes using user trajectories as sub-goals and text prompts as instructions.

What you can do with it

  • Reason over scenes at high and detailed levels
  • Align motion and text through frame alignment
  • Adapt real motion clips to synthetic terrain for variety
  • Follow text instructions to generate character actions

Application scenarios

  • A VR game generates motion that fits terrain and commands
  • An AR app builds navigation paths from natural language
  • A film team produces character motion in complex scenes

Pros & cons

✓ What we like

  • Real motion clips adapted to synthetic terrain add variety
  • Tested on several real-world scene datasets

! What to watch out for

  • Navigation depends on supplied trajectories and text prompts
  • Goal-centric normalisation is used to avoid clipping and floating

FAQ

What does it control?

The material describes character navigation in 3D scenes from trajectories and text prompts.

Which fields are named?

The material names virtual reality, augmented reality and game development.

Last reviewed: 2026-11-15

More AI office assistant tools

View all →

How we review