Visual concept personalization aims to transfer only specific image attributes, such as identity, expression, lighting, and style, into unseen contexts. However, existing methods rely on holistic embeddings from general-purpose image encoders, which entangle multiple visual factors and make it difficult to isolate a single attribute. This often leads to information leakage and incoherent synthesis. To address this limitation, we introduce Omni-Attribute, the first open-vocabulary image attribute encoder designed to learn high-fidelity, attribute-specific representations. Our approach jointly designs the data and model: (i) we curate semantically linked image pairs annotated with positive and negative attributes to explicitly teach the encoder what to preserve or suppress; and (ii) we adopt a dual-objective training paradigm that balances generative fidelity with contrastive disentanglement. The resulting embeddings prove effective for open-vocabulary attribute retrieval, personalization, and compositional generation, achieving state-of-the-art performance across multiple benchmarks.
@article{arxiv.2512.10955,
title = {Omni-Attribute: Open-vocabulary Attribute Encoder for Visual Concept Personalization},
author = {Tsai-Shien Chen and Aliaksandr Siarohin and Gordon Guocheng Qian and Kuan-Chieh Jackson Wang and Egor Nemchinov and Moayed Haji-Ali and Riza Alp Guler and Willi Menapace and Ivan Skorokhodov and Anil Kag and Jun-Yan Zhu and Sergey Tulyakov},
journal= {arXiv preprint arXiv:2512.10955},
year = {2026}
}