AnimaX - A 3D animation generation framework jointly developed by Beihang University, Tsinghua University, and others.

What is AnimaX?

AnimaX is a high-efficiency 3D animation generation framework developed by Beihang University in collaboration with Tsinghua University, the University of Hong Kong, and others. It combines the motion priors of a video diffusion model with a skeleton-based animation structure. The framework effectively transfers motion knowledge from videos to the 3D domain, supporting diverse joint meshes with arbitrary skeletal structures. AnimaX is based on representing 3D motion as multi-view, multi-frame 2D pose maps. Using a joint video pose diffusion model, combined with shared position encoding and modality-aware embedding, it ensures spatiotemporal alignment between video and pose sequences, achieving high-quality 3D animation generation. The framework's training dataset contains 160,000 binding sequences. AnimaX achieves industry-leading levels in generalization, motion fidelity, and efficiency, providing a scalable solution for class-agnostic 3D animation.

AnimaX's main functions

• Support for arbitrary skeletal structuresAnimaX supports processing 3D models with different skeletal topologies, suitable for various characters and objects, such as humans, animals, furniture, etc.
• Text-driven animationUsers specify animation content based on simple text descriptions, and AnimaX generates corresponding animation sequences based on the text prompts.
• Multi-view consistencyThe generated animation remains consistent across multiple perspectives, ensuring its continuity and realism when viewed from different angles.
• High-efficiency generationAnimaX is based on a feedforward method, which generates high-quality 3D animations in a short time, greatly improving the efficiency of animation production.

AnimaX's technical principles

• Joint video pose diffusion modelAnimaX represents 3D motion as multi-view, multi-frame 2D pose graphs. It supports powerful motion priors from video pose diffusion models, maintaining compatibility with 3D skeletal animation. By jointly generating video pose diffusion models, AnimaX can simultaneously generate videos and corresponding 2D pose sequences. A joint generation strategy ensures spatiotemporal alignment between video and pose.
• Shared position encoding and modality-aware embeddingTo ensure spatiotemporal alignment between video and pose sequences, AnimaX introduces shared positional encoding. This encoding mechanism allows the model to better understand and align spatial information across different modalities (RGB video and pose graph). Modality-aware embedding is used to distinguish between different modalities, ensuring the model correctly handles the differences between RGB video and pose graph.
• Multi-view consistencyAnimaX introduces a multi-view attention mechanism and camera conditional embedding to ensure that the generated videos remain consistent across multiple viewpoints. This allows the model to learn the spatial correspondences between different viewpoints, generating consistent multi-view videos.
• 3D motion reconstruction and animationThe generated multi-view pose sequence is converted into 3D joint positions based on triangulation. Inverse kinematics is then used to convert the 3D joint positions into joint angles, driving the animation of the 3D model.
• Training on large datasetsAnimaX was trained on a newly curated dataset containing 160,000 binding sequences. The data covers a variety of categories, such as humans, animals, and furniture, ensuring the model has good generalization ability.

AnimaX project address

• Project official websitehttps://anima-x.github.io/
• GitHub repositoryhttps://github.com/anima-x/anima-x
• arXiv technical paper: https://arxiv.org/pdf/2506.19851

AnimaX application scenarios

• Game developmentQuickly generate animations for characters and environments, improving development efficiency and content richness.
• Film and television productionUsed in animated films, visual effects, and the generation of animations for virtual characters to enhance visual effects.
• Virtual Reality (VR) and Augmented Reality (AR)It generates animations for virtual characters and dynamic environments, enhancing immersion and interactive experience.
• Advertising and MarketingCreate dynamic ads and product display animations to attract audience attention and increase engagement.
• Education and trainingGenerate animations for virtual experiments and simulation training to improve teaching and training effectiveness.