Oxygen-TryOn: Fashion-Native Foundation Model for Any-item Virtual Try-On

🔍 Overview

Oxygen-TryOn is a unified, open-source foundation model for any-item virtual try-on. Given one or more reference items — provided either as clean product shots or as in-the-wild photos of someone already wearing them — together with a single target subject image, the model synthesizes a photorealistic image of that subject wearing the referenced items, spanning virtually any fashion category: clothing, outerwear, accessories, footwear, bags, and beyond.

Unlike general-purpose image editors merely prompted for the task, Oxygen-TryOn is fashion-native: instead of treating try-on as mask-based inpainting, it reformulates it as a multi-reference, understanding-driven generation task, and is built specifically for try-on through a dedicated data engine and try-on-specific training. It accepts a variable number of references, composes multiple items in a single generation pass, and reasons holistically about layering and occlusion across full-body and half-body views, diverse poses, and non-standard subjects.

Under the hood, Oxygen-TryOn is built on the JoyAI-Image-Edit architecture and initialized from its pretrained weights, coupling a multimodal large language model (MLLM) for reference and instruction understanding with a multimodal diffusion transformer (MMDiT) for high-fidelity synthesis. It is trained with a three-stage recipe — continued pre-training (CPT), large-scale supervised fine-tuning (SFT), and reinforcement learning (RL) under a hybrid reward — and retains the general instruction-based editing ability of its foundation (e.g., pose change) within the same generation pass.

To our knowledge, Oxygen-TryOn is the first open-source system to deliver any-item, multi-reference try-on at this level of fidelity, achieving state-of-the-art consistency and realism that surpasses strong proprietary systems such as Nano Banana Pro, GPT-Image-2, and Seedream5 Lite, as well as leading open-source models such as FLUX.2.

✨ Key Features

• 🧥 Any item, any combination — garments, outerwear, accessories, shoes, bags, and more; from a single item to free multi-item outfits, with the model resolving layering and occlusion ("OOTD"-style full-outfit composition).
• 🖼️ Heterogeneous references — accepts both clean product shots and in-the-wild worn-on photos; full- or half-body subjects with a variable number of references.
• 🧍 Faithful preservation — keeps both the subject's identity and the referenced items' appearance intact.
• ✏️ Built-in editing — general instruction-based edits (e.g., pose change) within the same generation pass, with no second model or pass.
• 🎭 Cross-domain generalization — even dresses stylized 3D avatars, illustrated characters, statues, or posters while respecting the original style and geometry.
• 🏆 State-of-the-art single-item consistency & realism, surpassing strong proprietary systems (Nano Banana Pro, GPT-Image-2, Seedream5 Lite) and leading open-source models (FLUX.2).