WithEveryone: Unified Planning and Identity Grounding for Group Image Generation (arXiv 2608.20336)

Forum topic · 小凯 · 2026-08-22

Summary

WithEveryone is a unified framework for identity-preserving group image generation that can incorporate up to ten reference identities in a single image. The system injects each selected identity as an addressing token, predicts a structured identity-layout plan, and renders that plan as visual conditioning. Its core objective, Layout-Grounded ID Loss, directly supervises target identities using annotated face regions, avoiding unstable embedding-based face matching. A technique called ID Representation Forcing trains per-identity predictions before image synthesis. On benchmarks with disjoint identities, WithEveryone achieves the highest targeted contextual identity similarity, improving face similarity from GPT-Image-2's 0.462 to 0.499 while reducing copy-paste artifacts from 0.169 to 0.055. It covers 97.3% of requested identities with a duplication rate of only 2.8%. Paper: arXiv 2608.20336, by Hengyuan Xu, Qixun Wang, Yiji Cheng, and Miles Yang (computer vision).

Paper Overview

Field: Computer Vision (CV) Authors: Hengyuan Xu, Qixun Wang, Yiji Cheng, Miles Yang Published: 2026-08-22 arXiv: 2608.20336

Abstract (English)

Identity-preserving image generation becomes increasingly unreliable when scenes need to include multiple specified individuals. Beyond preserving each identity, the model must bind each reference to distinct people and locations. This paper proposes WithEveryone, a unified framework for generating group images containing up to ten reference identities.

WithEveryone:

Injects each selected identity as an addressing token
Predicts a structured identity-layout plan
Renders the plan as visual conditioning

Its core objective, Layout-Grounded ID Loss, directly supervises target identities using annotated face regions, avoiding unstable embedding-based face matching. Additionally, ID Representation Forcing trains per-identity predictions before image synthesis.

Results

On benchmarks with disjoint identities, WithEveryone achieves the highest targeted contextual identity similarity:

| Metric | Baseline (GPT-Image-2) | WithEveryone | |---|---|---| | Face similarity | 0.462 | 0.499 | | Copy-paste artifacts | 0.169 | 0.055 | | Requested identity coverage | — | 97.3% | | Duplication rate | — | 2.8% |

Full paper: arXiv 2608.20336

--- *Auto-collected on 2026-08-22*