# UniCustom: Unified Visual Conditioning for Multi-Reference Image Generation arXiv: https://arxiv.org/abs/2605.12088 ยท HTML: https://mlsanity.com/paper/2605.12088/ ## Key facts - Cited by 2 papers in the map (data through 14 Oct 2026). - Builds on 39 of its 43 references that are in the map. - First posted on arXiv 12 May 2026. ## Builds on (most cited first, 30 of 39) - [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014) - [A Style-Based Generator Architecture for Generative Adversarial Networks](https://mlsanity.com/paper/1812.04948/) (2018) - [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025) - [IP-Adapter: Text Compatible Image Prompt Adapter for Text-to-Image Diffusion Models](https://mlsanity.com/paper/2308.06721/) (2023) - [Qwen-Image Technical Report](https://mlsanity.com/paper/2508.02324/) (2025) - [FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space](https://mlsanity.com/paper/2506.15742/) (2025) - [Emerging Properties in Unified Multimodal Pretraining](https://mlsanity.com/paper/2505.14683/) (2025) - [BLIP-Diffusion: Pre-trained Subject Representation for Controllable Text-to-Image Generation and Editing](https://mlsanity.com/paper/2305.14720/) (2023) - [Step1X-Edit: A Practical Framework for General Image Editing](https://mlsanity.com/paper/2504.17761/) (2025) - [BLIP3-o: A Family of Fully Open Unified Multimodal Models-Architecture, Training and Dataset](https://mlsanity.com/paper/2505.09568/) (2025) - [Open-Sora Plan: Open-Source Large Video Generation Model](https://mlsanity.com/paper/2412.00131/) (2024) - [Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer](https://mlsanity.com/paper/2511.22699/) (2025) - [OmniControl: Control Any Joint at Any Time for Human Motion Generation](https://mlsanity.com/paper/2310.08580/) (2023) - [ImgEdit: A Unified Image Editing Dataset and Benchmark](https://mlsanity.com/paper/2505.20275/) (2025) - [Less-to-More Generalization: Unlocking More Controllability by In-Context Generation](https://mlsanity.com/paper/2504.02160/) (2025) - [MS-Diffusion: Multi-subject Zero-shot Image Personalization with Layout Guidance](https://mlsanity.com/paper/2406.07209/) (2024) - [Kosmos-G: Generating Images in Context with Multimodal Large Language Models](https://mlsanity.com/paper/2310.02992/) (2023) - [Echo-4o: Harnessing the Power of GPT-4o Synthetic Images for Improved Image Generation](https://mlsanity.com/paper/2508.09987/) (2025) - [LongCat-Image Technical Report](https://mlsanity.com/paper/2512.07584/) (2025) - [HunyuanCustom: A Multimodal-Driven Architecture for Customized Video Generation](https://mlsanity.com/paper/2505.04512/) (2025) - [DreamO: A Unified Framework for Image Customization](https://mlsanity.com/paper/2504.16915/) (2025) - [Human-Art: A Versatile Human-Centric Dataset Bridging Natural and Artificial Scenes](https://mlsanity.com/paper/2303.02760/) (2023) - [UniWorld-V1: High-Resolution Semantic Encoders for Unified Visual Understanding and Generation](https://mlsanity.com/paper/2506.03147/) (2025) - [Pico-Banana-400K: A Large-Scale Dataset for Text-Guided Image Editing](https://mlsanity.com/paper/2510.19808/) (2025) - [XVerse: Consistent Multi-Subject Control of Identity and Semantic Attributes via DiT Modulation](https://mlsanity.com/paper/2506.21416/) (2025) - [TokenVerse: Versatile Multi-concept Personalization in Token Modulation Space](https://mlsanity.com/paper/2501.12224/) (2025) - [NoHumansRequired: Autonomous High-Quality Image Editing Triplet Mining](https://mlsanity.com/paper/2507.14119/) (2025) - [USO: Unified Style and Subject-Driven Generation via Disentangled and Reward Learning](https://mlsanity.com/paper/2508.18966/) (2025) - [Bee: A High-Quality Corpus and Full-Stack Suite to Unlock Advanced Fully Open MLLMs](https://mlsanity.com/paper/2510.13795/) (2025) - [CINEMA: Coherent Multi-Subject Video Generation via MLLM-Based Guidance](https://mlsanity.com/paper/2503.10391/) (2025) ## Built on by (most cited first, 2 of 2) - [RefRoute: Decoupling Conditioning Cost from References via Compact Residual Conditioning and Spatial Routing](https://mlsanity.com/paper/2610.07720/) (2026) - [From Prompting to Composing: A Spatial Canvas Interface for Poster Generation](https://mlsanity.com/paper/2610.12230/) (2026) ## Abstract (arXiv) Multi-reference image generation aims to synthesize images from textual instructions while faithfully preserving subject identities from multiple reference images. Existing VLM-enhanced diffusion models commonly rely on decoupled visual conditioning: semantic ViT features are processed by the VLM for instruction understanding, whereas appearance-rich VAE features are injected later into the diffusion backbone. Despite its intuitive design, this separation makes it difficult for the model to associate each semantically grounded subject with visual details from the correct reference image. As a result, the model may recognize which subject is being referred to, but fail to preserve its identity and fine-grained appearance, leading to attribute leakage and cross-reference confusion in complex multi-reference settings. To address this issue, we propose UniCustom, a unified visual conditioning framework that fuses ViT and VAE features before VLM encoding. This early fusion exposes the VLM to both semantic cues and appearance-rich details, enabling its hidden states to jointly encode the referred subject and corresponding visual appearance with only a lightweight linear fusion layer. To learn such unified representations, we adopt a two-stage training strategy: reconstruction-oriented pretraining that preserves reference-specific appearance details in the fused hidden states, followed by supervised finetuning on single- and multi-reference generation tasks. We further introduce a slot-wise binding regularization that encourages each image slot to preserve low-level details of its corresponding reference, thereby reducing cross-reference entanglement. Experiments on two multi-reference generation benchmarks demonstrate that UniCustom consistently improves subject consistency, instruction following, and compositional fidelity over strong baselines.