[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-75467":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":12,"stars7d":16,"stars30d":17,"stars90d":15,"forks30d":15,"starsTrendScore":16,"compositeScore":18,"rankGlobal":9,"rankLanguage":9,"license":9,"archived":19,"fork":19,"defaultBranch":20,"hasWiki":19,"hasPages":19,"topics":21,"createdAt":9,"pushedAt":9,"updatedAt":22,"readmeContent":23,"aiSummary":24,"trendingCount":15,"starSnapshotCount":15,"syncStatus":14,"lastSyncTime":25,"discoverSource":26},75467,"MACE-Dance","AMAP-ML\u002FMACE-Dance","AMAP-ML","[SIGGRAPH 2026] MACE-Dance: Motion-Appearance Cascaded Experts for Music-Driven Dance Video Generation ",null,"Python",104,1,86,2,0,3,18,0.9,false,"main",[],"2026-06-12 02:03:34","\u003Cp align=\"center\">\n  \u003Ch1 align=\"center\">🎵 MACE-Dance\u003C\u002Fh1>\n  \u003Ch3 align=\"center\">Motion-Appearance Cascaded Experts for Music-Driven Dance Video Generation\u003C\u002Fh3>\n\u003Cp align=\"center\">\n    Kaixing Yang\u003Csup>1\u003C\u002Fsup> ·\n    Jiashu Zhu\u003Csup>2,*\u003C\u002Fsup> ·\n    Xulong Tang\u003Csup>5\u003C\u002Fsup> ·\n    Ziqiao Peng\u003Csup>1\u003C\u002Fsup> ·\n    Xiangyue Zhang\u003Csup>4\u003C\u002Fsup>\n    \u003Cbr>\n    Puwei Wang\u003Csup>1,†\u003C\u002Fsup> ·\n    Jiahong Wu\u003Csup>2,†\u003C\u002Fsup> ·\n    Xiangxiang Chu\u003Csup>2\u003C\u002Fsup> ·\n    Hongyan Liu\u003Csup>3,†\u003C\u002Fsup> ·\n    Jun He\u003Csup>1\u003C\u002Fsup>\n    \u003Cbr>\u003Cbr>\n    \u003Csup>1\u003C\u002Fsup>Renmin University of China  \n    \u003Csup>2\u003C\u002Fsup>AMap, Alibaba  \n    \u003Csup>3\u003C\u002Fsup>Tsinghua University  \n    \u003Csup>4\u003C\u002Fsup>Wuhan University  \n    \u003Csup>5\u003C\u002Fsup>Malou Tech Inc\n    \u003Cbr>\u003Cbr>\n    \u003Csup>*\u003C\u002Fsup>Project Leader  \n    \u003Csup>†\u003C\u002Fsup>Corresponding Authors\n\u003C\u002Fp>\n\n  \u003Cp align=\"center\">\n    \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2512.18181\">\n      \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-MACE--Dance-green\" alt=\"Paper\">\n    \u003C\u002Fa>\n    \u003Ca href=\"https:\u002F\u002Fsun-happy-ykx.github.io\u002FMACE-Dance\u002F\">\n      \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FProject_Page-MACE--Dance-blue\" alt=\"Project Page\">\n    \u003C\u002Fa>\n    \u003Ca href=\"#\">\n      \u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FConference-SIGGRAPH%202026-orange\" alt=\"Conference\">\n    \u003C\u002Fa>\n  \u003C\u002Fp>\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"teaser.png\" width=\"90%\" alt=\"MACE-Dance teaser\">\n\u003C\u002Fp>\n\n\u003Cp align=\"center\">\n  \u003Cem>\n    MACE-Dance is a cascaded expert framework for music-driven dance video generation,\n    explicitly decoupling motion generation and appearance synthesis to produce\n    kinematically plausible, artistically expressive, and visually coherent dance videos.\n  \u003C\u002Fem>\n\u003C\u002Fp>\n\n---\n\n## ✨ Overview\n\n**MACE-Dance** is the official PyTorch implementation of the SIGGRAPH 2026 paper:\n\n> **MACE-Dance: Motion-Appearance Cascaded Experts for Music-Driven Dance Video Generation**\n\nMusic-driven dance video generation is challenging because it requires simultaneously modeling:\n\n- **Motion quality**: kinematically plausible and artistically expressive dance motion\n- **Appearance quality**: high-fidelity visual synthesis with strong spatiotemporal consistency\n\nTo address this, **MACE-Dance** decomposes the task into two cascaded experts:\n\n- **🕺 Motion Expert**: generates music-aligned **3D dance motion**\n- **🎨 Appearance Expert**: synthesizes the final **dance video** conditioned on motion and reference appearance\n\nInstead of using 2D keypoints as the intermediate representation, MACE-Dance adopts **3D SMPL motion**, which provides better spatial fidelity, cleaner supervision, and stronger robustness for downstream video synthesis.\n\n---\n\n## 🧩 Repository Structure\n\n```bash\nMACE-Dance\u002F\n├── Expert-Motion\u002F           # Motion Expert: music-to-3D dance motion\n├── Expert-Appearance\u002F       # Appearance Expert: motion-guided video synthesis\n├── Evaluation-Motion\u002F       # Motion-dimension evaluation\n├── Evaluation-Appearance\u002F   # Appearance-dimension evaluation\n├── teaser.png\n└── README.md\n```\n---\n\n## 📚 MA-Data Dataset\n\nWe provide **MA-Data**, a large-scale dataset for music-driven dance video generation, containing **~70K video clips** spanning **116 hours** across **20+ dance genres**. Please refer to the [dataset page](https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002FGD-ML\u002FMACE-Dance) for more details.\n\n\n---\n\n## 🏋️ Model Weights\n\nThe source code for **MACE-Dance** is fully open-source. For the model weights of the Appearance Expert, please visit the link below to request access or download:\n\n👉 **[Click here to access MACE-Dance Model Weights](https:\u002F\u002Fhuggingface.co\u002FGD-ML\u002FMACE-Dance)**\n\n---\n\n## 📏 Evaluation Protocol\n\nWe provide a motion–appearance evaluation protocol for music-driven dance video generation, including motion quality assessment based on ViTPose keypoints and appearance quality assessment based on VBench. Please refer to `Evaluation-Motion` and `Evaluation-Appearance` for details.\n\n---\n\n## 📌 Notes\n\n- The repository is organized into **expert modules** and **evaluation modules**.\n- Please check the subfolder READMEs for environment setup, inference, and evaluation details.\n- Some released example files are for demonstration only; please replace them with your own predictions \u002F ground-truth files during evaluation.\n\n---\n\n## 📄 Citation\n\nIf you find this project useful, please consider citing our paper:\n\n```bibtex\n@inproceedings{yang2026macedance,\n  title={MACE-Dance: Motion-Appearance Cascaded Experts for Music-Driven Dance Video Generation},\n  author={Yang, Kaixing and Zhu, Jiashu and Tang, Xulong and Peng, Ziqiao and Zhang, Xiangyue and Wang, Puwei and Wu, Jiahong and Chu, Xiangxiang and Liu, Hongyan and He, Jun},\n  booktitle={Proceedings of the ACM SIGGRAPH Conference},\n  year={2026}\n}\n```\n\n---\n\n## 🙏 Acknowledgement\n\nThis work was supported in part by the National Nature Science Foundation of China under Grants **62436010**, **72572090**, **62572474**, and **62172421**, and in part by the Tsinghua University School of Economics and Management Research Grant.\n","MACE-Dance 是一个用于音乐驱动的舞蹈视频生成的框架，通过级联专家模型分别处理动作生成和外观合成。其核心功能包括使用3D SMPL运动数据生成与音乐同步的高质量舞蹈动作，并基于这些动作生成视觉上连贯且具有高保真度的舞蹈视频。该项目采用Python开发，适合于需要创建高质量、艺术性强且与背景音乐相匹配的舞蹈内容的应用场景，如虚拟表演制作、游戏动画设计等。","2026-06-11 03:52:54","CREATED_QUERY"]