WithEveryone targets group image identity grounding
The framework plans identity and layout to generate group images with up to 10 referenced people.
Why it matters
Group image generation remains difficult because models must preserve multiple identities while binding each one to the right person and location. WithEveryone suggests explicit identity-layout planning can improve reliability for multi-person image workflows.
The key points
- 1.Generates group images with up to 10 reference identities.
- 2.Improved face similarity versus GPT-Image-2 on the benchmark.
- 3.Reduced copy-paste artifacts and duplicate identities.
Researchers introduced WithEveryone, a unified framework for group image generation with up to 10 reference identities. The method injects each identity as an addressed token, predicts a structured identity-layout plan, and renders that plan as a visual condition. On an identity-disjoint benchmark, it improved face similarity from 0.462 for GPT-Image-2 to 0.499 and reduced copy-paste artifacts from 0.169 to 0.055.
⚡ Try this today
Read the paper before building multi-person identity-preserving image generation pipelines.
Sources & original reporting
This brief summarizes and links to reporting from the publishers below.
Enjoyed this brief? Get the next one in your inbox.
More in Research
Study finds AI signs in a third of post-ChatGPT web pages
Research cited by TechCrunch says AI models are shaping much of the new web.
OpenAI math solutions stir debate among mathematicians
The Verge reports that OpenAI's work on longstanding math problems has raised questions for the field.
Preprint proposes spectral neuron for interpretable ML
The model uses learned symmetric matrices and eigenvalues to add nonlinear expressiveness while retaining transparency.