3D-Muse
An open-source end-to-end single-image 3D digital human generation model from Tencent Hunyuan team. It performs 3D reconstruction, skeleton rigging, material texturing, and expression driving configuration from a single image, and outputs complete assets exportable to Blender, Unity, and Unreal Engine.
Timeline (last 90 days)
The Tencent Hunyuan team has open-sourced 3D-Muse, which can generate a drivable digital human with expressions and lip-sync from a single character illustration or photo.
3D-Muse uses an end-to-end automated generation pipeline: after a user uploads a single-view image, the system completes 3D reconstruction, skeleton rigging, material texturing and expression-driving configuration within minutes.
Results generated by 3D-Muse can be exported to Blender, Unity or Unreal Engine, with rigged skeletons and expression morph targets preserved.
3D-Muse uses generative priors to complete back and side-face textures, and aligns the lighting of the original image through a differentiable rendering module, making the reconstruction highly consistent with the original image in the frontal view.
3D-Muse provides an audio-driven lip-sync animation interface for the generated digital human: input speech and lip-shape keyframes are generated automatically, with no manual lip-syncing needed.
The 3D-Muse team provides complete environment configuration scripts and example projects in the open-source repository, lowering the barrier to reproduction.
The current version of 3D-Muse has certain requirements for input image quality: clear, evenly lit frontal character illustrations work best, while complex occlusion or low-resolution inputs may produce texture stretching.
Stability of facial deformation under exaggerated expressions and multi-view consistency remain common challenges in single-image 3D reconstruction, and 3D-Muse has not fully solved them.
3D-Muse compresses modeling, topology and rigging work that previously took hours or even days down to the click of a button.
3D-Muse is suited to virtual streamers and short-video creators, game and animation teams, education and training scenarios, and researchers as a reference implementation.
All Facts (10)
3D-Muse is suited to virtual streamers and short-video creators, game and animation teams, education and training scenarios, and researchers as a reference implementation.
50%Unverified3D-Muse uses an end-to-end automated generation pipeline: after a user uploads a single-view image, the system completes 3D reconstruction, skeleton rigging, material texturing and expression-driving configuration within minutes.
50%UnverifiedResults generated by 3D-Muse can be exported to Blender, Unity or Unreal Engine, with rigged skeletons and expression morph targets preserved.
50%Unverified3D-Muse uses generative priors to complete back and side-face textures, and aligns the lighting of the original image through a differentiable rendering module, making the reconstruction highly consistent with the original image in the frontal view.
50%Unverified3D-Muse provides an audio-driven lip-sync animation interface for the generated digital human: input speech and lip-shape keyframes are generated automatically, with no manual lip-syncing needed.
50%UnverifiedThe 3D-Muse team provides complete environment configuration scripts and example projects in the open-source repository, lowering the barrier to reproduction.
50%UnverifiedThe current version of 3D-Muse has certain requirements for input image quality: clear, evenly lit frontal character illustrations work best, while complex occlusion or low-resolution inputs may produce texture stretching.
50%UnverifiedStability of facial deformation under exaggerated expressions and multi-view consistency remain common challenges in single-image 3D reconstruction, and 3D-Muse has not fully solved them.
50%Unverified3D-Muse compresses modeling, topology and rigging work that previously took hours or even days down to the click of a button.
50%UnverifiedThe Tencent Hunyuan team has open-sourced 3D-Muse, which can generate a drivable digital human with expressions and lip-sync from a single character illustration or photo.
50%