AlayaWorld — Weights

Pretrained inference weights for AlayaWorld: Long-Horizon and Playable Video World Generation (Alaya Lab, Shanda AI Research Tokyo).

• Code: https://github.com/AlayaLab/AlayaWorld
• Technical report: https://arxiv.org/abs/2607.06291

Files

File	Description
merged_infer.safetensors	Single bundle: DiT + VAE + text-encoder + history-encoder. Place under checkpoints/ per the code repo's README.

The text encoder (Gemma-3) and the spatial-memory depth model (Depth-Anything-3) are not included here — download them from their original sources as described in the code repository's README.

License

This project is based on LTX-2 by Lightricks Ltd. merged_infer.safetensors is fine-tuned from the LTX-2.3-22B base and is a derivative of LTX-2.3; accordingly it is released under the LTX-2 Community License Agreement (https://github.com/Lightricks/LTX-2/blob/main/LICENSE), not Apache 2.0. All original LTX-2 copyright and attribution notices are retained.

For academic research and non-commercial use only. For commercial use of LTX-2 or its derivatives, contact Lightricks Ltd.; entities with ≥ $10M annual revenue require a commercial license from Lightricks.

Citation

@article{team2026alayaworld,
  title={AlayaWorld: Long-Horizon and Playable Video World Generation},
  author={Team, AlayaWorld and Zhang, Kaipeng and Li, Chuanhao and Zhan, Yifan and Ge, Yongtao and Yin, Yuanyang and others},
  journal={arXiv preprint arXiv:2607.06291},
  year={2026}
}