[ComfyUI] Krea 2 Dual-Image Style Fusion (Ultimate Visual Extractor)

Description: 🚀 Run this workflow for FREE on Runninghub:

🎥 Watch the full YouTube tutorial to see it in action:

💬 Join Discord for the System Prompt Templates:

Overview Where Krea 2 truly shines is not just in plain text-to-image, but in its insane ability to extract and fuse "visual flavors." Unlike traditional editing models that just swap backgrounds or mask subjects, Krea 2 breaks down reference images into dimensions like style, texture, lighting, and mood, and regenerates an entirely fused composition.

If you want to borrow the lighting of Image A and combine it with the character pose of Image B, this is the workflow you need.

⚙️ Core Features:

• Dual-Vector Extraction: Feed the workflow two distinct reference images. You can assign one to control the main subject/pose, and the other to dictate the environmental lighting, color tone, or brushstrokes.

• LLM Reverse-Engineering: Uses an online Large Language Model (Cannot be used offline

  )to reverse-engineer the style of your references and generate a highly stable fusion prompt .（you can change it to local LLM）

• True Style Blending: Perfect for creating personalized art styles by mixing elements from multiple sources.

⚠️ Creator's Disclaimer (Subject Drift): Because this model completely regenerates the fused image, you might experience slight subject drift (especially with faces). I highly recommend running the output through a standard Face-Swap or detailing workflow afterward if you need strict character consistency!