Skip to content

DreamActor-M1

Animate a human image with fine-grained control

Visit official site

What DreamActor-M1 is

The framework is described as a diffusion transformer (DiT) human animation framework with fine-grained holistic control, multi-scale adaptation and long-term temporal consistency. The model uses hybrid guidance to generate expressive and realistic human videos with high fidelity and identity preservation.

What you can do with it

  • Control facial expressions and body motion with an implicit face representation, a 3D head sphere and a 3D body skeleton
  • Adapt to varied body poses and image resolutions through progressive training
  • Keep unobserved regions consistent over complex motions
  • Extend to audio-driven facial animation with multilingual lip sync

Application scenarios

  • Produce high-quality human character animation in animated film work
  • Give game characters smooth motion in game development
  • Make short videos for social media content creation

Getting started

  • Prepare a reference image and driving video frames
  • Feed the reference image and video frames into the model
  • Set hybrid guidance parameters to adjust face and body motion
  • Run the model to generate the target animation video

Pros & cons

✓ What we like

  • High fidelity with identity preservation
  • Supports portrait and full-body views at different resolutions

! What to watch out for

  • Motion transfer needs reference images and driving video frames
  • Hybrid guidance parameters have to be set for face and body motion

FAQ

What is the framework based on?

The material describes a diffusion transformer (DiT) framework for human image animation.

Can it handle only part of a motion?

The feature list states partial motion transfer, such as facial expressions and head motion, is supported.

Last reviewed: 2026-11-15

More AI video generation tools

View all →

How we review