We are the HumanAIGC Team at Tongyi, Alibaba. We focus on the understanding and generation of human-centric content.

• WanToFight: Real-Time Generative Game Engine for Multi-Player Combat Interaction. Paper Project
• Wan-Dancer: A Hierarchical Framework for Minute-scale Coherent Music-to-Dance Generation. Paper, Project, Code
• UCM: Unifying Camera Control and Memory with Time-aware Positional Encoding Warping for World Models. Paper, Project, Code
• SyncAnyone: Implicit Disentanglement via Progressive Self-Correction for Lip-Syncing in the wild. Paper, Project
• Knot Forcing: Taming Autoregressive Video Diffusion Models for Real-time Infinite Interactive Portrait Animation. Paper, Project, Code
• Co-speech Gesture Video Generation via Motion-Based Graph Retrieval. Paper
• Wan-Animate: Unified Character Animation and Replacement with Holistic Replication. Paper, Project, Code
• Wan-S2V: Audio-Driven Cinematic Video Generation. Paper, Project, Code
• MirrorMe: Towards Realtime and High Fidelity Audio-Driven Halfbody Animation. Paper
• SwapAnyHead: Controllable and Expressive One-Shot Video Head Swapping. Project, Paper
• FaceTimelineControl: Exploring Timeline Control for Facial Motion Generation. Project, Paper
• MotionRAG-Diff: A Retrieval-Augmented Diffusion Framework for Long-Term Music-to-Dance Generation. Project, Paper
• OmniTalker: Real-Time Text-Driven Talking Head Generation with In-Context Audio-Visual Style Replication. Project, Paper, Demo
• ChatAnyone: Stylized Real-time Portrait Video Generation with Hierarchical Motion Diffusion Model. Project, Paper
• LiteAvatar: a audio2face model for realtime 2D chat avatar. Code
• Animate Anyone 2: High-Fidelity Character Image Animation with Environment Affordance. Project, Paper
• EMO2: End-Effector Guided Audio-Driven Avatar Video Generation. Project, Paper
• EMO: Emote Portrait Alive - Generating Expressive Portrait Videos with Audio2Video Diffusion Model under Weak Conditions. Project, Paper
• Outfit Anyone: Ultra-high quality virtual try-on for Any Clothing and Any Person. Project, Paper, Demo
• Animate Anyone: Consistent and Controllable Image-to-Video Synthesis for Character Animation. Project, Paper
• VividTalk: One-Shot Audio-Driven Talking Head Generation Based on 3D Hybrid Prior. Project, Paper
• DanceMeld: Unraveling Dance Phrases with Hierarchical Latent Codes for Music-to-Dance Synthesis. Paper
• Cloth2Tex: A Customized Cloth Texture Generation Pipeline for 3D Virtual Try-On. Project, Paper, Code