Bring everyone into the frame.
WithEveryone generates coherent group images from five to ten reference identities while preserving who is who.
An open-source release is on the way!
The research version presented in our paper is built on a foundation model whose licensing terms do not allow us to release its checkpoint.
To provide the community with an open alternative, we are actively training a new version on a foundation model that supports open release.
Code and checkpoints will be shared once the new version is readyโstay tuned!
Address every identity. Each reference becomes a dedicated identity token.
Plan before generation. Identity-aware layout reasoning organizes every person in the scene.
Supervise the right face. Layout-Grounded ID Loss applies identity supervision to the intended region.
@article{xu2026witheveryone,
title={WithEveryone: Unified Planning and Identity Grounding for Group Image Generation},
author={Xu, Hengyuan and Wang, Qixun and Cheng, Yiji and Yang, Miles and Zhong, Zhao and Cheng, Wei and Ma, Xingjun and Jiang, Yu-Gang},
journal={arXiv preprint arXiv:2608.20336},
year={2026}
}