USO Unified Style and Subject-Driven Generation Now Available in ComfyUI

ByteDance's USO model brings unified style and subject generation capabilities to ComfyUI

We’re excited to announce that USO (Unified Style-Subject Optimized), ByteDance’s unified style and subject-driven generation model, is now natively supported in ComfyUI!

USO is developed by ByteDance’s UXO Team and built on the FLUX architecture, representing the first model to successfully unify style-driven and subject-driven generation tasks. The model achieves both style similarity and subject consistency through disentangled learning and style reward learning (SRL), delivering SOTA performance among open-source models.

Model Highlights

• Unified Framework: First model to combine style transfer and subject consistency in one framework

• Three Generation Modes: Subject-driven, style-driven, and combined style-subject generation

• Multi-Style Support: Blend multiple artistic styles for unique effects

• Layout Control: Preserve original composition or transform layouts as needed

Getting Started

1. Update ComfyUI to the latest version

2. Click the templates icon on the sidebar → Flux → Flux.1 Dev USO Reference Image Generation

3. Download the model as guided by the pop-up dialog

4. Follow the guide in the template, then run the workflow

Examples

Example 1

Input

Content reference

Prompt: A European girl with a heartfelt smile, her hands cradling a delicate bouquet. She is immersed in a vast, endless field of blooming flowers under a perfect summer sky.

Example 2

Input

Content reference

Style reference

Prompt: A little puppy fell asleep in the forest

Example 3

Input

Prompt: a child's room

Example 4

Input

Prompt:A man dressed fashionably stands on the forest.

Check our blog for more details: https://docs.comfy.org/tutorials/flux/flux-1-uso
As always, enjoy creating!

​