📖 Overview

Note

AniLayer2D is a layered image dataset tailored for Live2D models, featuring multi-layer assets, semantic label, and descriptive captions. Originally inspired by the emergence of LayerDiffusion, we recognized the potential of such technology in Live2D production. Constrained by computational resources at the time, we chose to contribute by building this dataset. Although completed in February 2025 , the release was delayed. The recent launch of Qwen-Image-Layered has reignited our belief that fully automated Live2D generation is on the horizon. Consequently, we have organized and released this dataset to the community. If you find this helpful, please give us a Star ⭐!

🛠️ TODO List

• Upload data labeling results
• Upload part image dataset
• Upload data using sample code

📦 Dataset Format

AniLayer2D_Data_Root
    ├── <model_name>           
    │   ├── <model_type>
    │   │   ├── part
    │   │   ├── label_inf.json

• 🧾 label_inf.json: A json file containing the category of each layer and the mapping of image sequence numbers, label name and character caption.

🎁 Acknowledgements

Model

• live2d: https://github.com/imuncle/live2d
• live2d-models: https://github.com/oh-my-live2d/live2d-models
• live2d-model-assets: https://github.com/zenghongtu/live2d-model-assets

Code

• live2d-v2: https://github.com/EasyLive2D/live2d-v2 Greatly appreciated for the support provided by Arkueid !!!

☎️ Community & Contact

This dataset was created on a whim, with limited knowledge in this field. Future plans include integrating multi-layer processing for moc/moc3/spine data. If interested or experienced, feel free to reach out. Guidance in this area would be greatly appreciated.

• E-mail: Corporate Email / Google Email