Skip to content

Story-to-Motion

Turn a long story into controllable character motion

Visit official site

What Story-to-Motion is

The framework is described as a task that takes a story and generates motions and trajectories matching the text, using a large language model as a text-driven motion scheduler.

What you can do with it

  • Synthesise unlimited controllable character animation from long text
  • Extract a series of text-position pairs from long text with an LLM
  • Retrieve motions with classic motion matching, motion semantics and trajectory constraints
  • Use a progressive masked transformer to address foot sliding and unnatural poses

Application scenarios

  • Generate character animation from game plot text in game development
  • Generate character motions from scripts in film and TV production
  • Let animation designers synthesise character animation from text

Model notes

  • The LLM serves as a text-driven motion scheduler
  • Transition problems named include unnatural poses and foot sliding
  • The material reports results on trajectory following, temporal action composition and motion blending
  • The source states it outperformed earlier motion synthesis methods

Pros & cons

✓ What we like

  • Story text drives both motions and trajectories
  • Motion retrieval combines classic matching with semantics and trajectory constraints

! What to watch out for

  • Motion scheduling depends on a large language model reading long text
  • Transition problems are addressed through a dedicated masked transformer

FAQ

What does it take as input?

The material states it accepts a story and generates motions and trajectories matching the text.

Which subtasks are evaluated?

Trajectory following, temporal action composition and motion blending are named.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review