OmniGen2: Image Edit

Introduction

OmniGen2 is a powerful and efficient unified multimodal generative model with a total parameter size of about 7B (3B text model + 4B image generation model). Unlike OmniGen v1, OmniGen2 adopts an innovative dual-path Transformer architecture with completely independent text autoregressive model and image diffusion model to achieve parameter decoupling and specialized optimization.

Model highlights

• Visual understanding: Inherits the powerful image content interpretation and analysis capabilities of the Qwen-VL-2.5 base model
• Text image generation: Create high-fidelity and beautiful images from text prompts
• Instruction-guided image editing: Perform complex, instruction-based image modifications, achieving state-of-the-art performance among open source models
• Context generation: Versatile ability to process and flexibly combine diverse inputs (including people, reference objects, and scenes) to produce novel and coherent visual outputs

Technical features

• Dual-path architecture: Based on Qwen 2.5 VL (3B) text encoder + independent diffusion Transformer (4B)
• Omni-RoPE position encoding: Supports multi-image spatial positioning and identity differentiation
• Parameter decoupling design: Avoids the negative impact of text generation on image quality
• Supports complex text understanding and image understanding
• Controllable image generation and editing
• Excellent detail preservation capabilities
• Unified architecture supports a variety of image generation tasks
• Text generation capabilities: Can generate clear text content in images

Recommended machine：Large-PRO

Workflow Overview

How to use this workflow

Step 1: Load Image

Use **Ctrl + B** to switch all nodes in Bypass mode to normal mode to enable the second image input.

Step 2: Input the Prompt

Step 3: Adjust Image parameters

Step 4: Get Image