Skip to content

DreamWaltz-G

Text-driven 3D avatars with expressive full-body animation

Visit official site

What DreamWaltz-G is

The framework is described as a text-driven system for 3D avatars and expressive full-body animation, combining skeleton-guided score distillation with a hybrid 3D Gaussian avatar representation.

What you can do with it

  • Create a 3D avatar from a text description
  • Animate the avatar with expressive full-body motion
  • Render results in real time

Getting started

  • Visit the DreamWaltz-G project page
  • Read the introduction and the related paper
  • Check the code repository for needed software and code
  • Set up the environment and dependencies from the documentation

Pros & cons

✓ What we like

  • Viewpoint and pose consistency is improved by skeleton control
  • Video reenactment and multi-subject scene composition are listed as uses

! What to watch out for

  • The approach relies on a 2D diffusion model
  • Setup requires code, software and dependencies from the project

FAQ

What are the two core ideas?

The material names skeleton-guided score distillation and a hybrid 3D Gaussian avatar representation.

What issues does it address?

The material lists multiple faces, extra limbs and blurring.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review