[ECCV 2026] InsertAnywhere: Geometrically Grounded and Optics-Aware Video Object Insertion

Hoiyeong Jin†, Hyojin Jang†, Junha Hyung†, Jeongho Kim†, Kinam Kim, Dongjin Kim, Huijin Choi, Hyeonji Kim, and Jaegul Choo

KAIST AI, SK Telecom
ECCV 2026. († indicates equal contribution)

🎬 Teaser Video

InsertAnywhere_Final.mp4

📝 Citation

If you use this dataset or code in your research, please cite our paper:

@misc{jin2026insertanywheregeometricallygroundedopticsaware,
      title={InsertAnywhere: Geometrically Grounded and Optics-Aware Video Object Insertion}, 
      author={Hoiyeong Jin and Hyojin Jang and Junha Hyung and Jeongho Kim and Kinam Kim and Dongjin Kim and Huijin Choi and Hyeonji Kim and Jaegul Choo},
      year={2026},
      eprint={2512.17504},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2512.17504}, 
}