1. Workflow Overview

This workflow leverages Flux-Kontext engine for professional-grade image manipulation:

• Localized Editing: Modify hair color, clothing, etc.

• Multi-version Comparison: Generate and concatenate variants (e.g., red vs. blue hair).

• Bilingual Support: Auto-translate prompts (Chinese ↔ English).

• Aspect Ratio Lock: Preserve original dimensions (e.g., 2:3).

Use Cases:

• E-commerce model image retouching.

• Old photo posture adjustment.

• Design variant comparisons.

2. Core Models & Tech

3. Key Nodes & Installation

Dependencies:

• Download Flux-Kontext models to models/flux_kontext/.

• Recommended VRAM ≥8GB for resolutions >1500px.

4. Workflow Structure

1. Input Group:

   • Load source image (e.g., 055.jpg).

   • Enter prompts (e.g., "change hair to red").

2. Processing Group:

   • FluxKontextPro: Simple edits (e.g., color change).

   • FluxKontextMax: Complex edits (e.g., standing pose).

3. Output Group:

   • Single edited image (SaveImage).

   • Comparison grid (ImageConcatMulti).

5. Inputs & Outputs

• Inputs:

  • Required: 1 source image (e.g., 1056x1584 portrait).

  • Optional:

    • Prompts (bilingual supported).

    • Seed (for reproducible results).

    • Aspect ratio (e.g., 2:3).

• Outputs:

  • Edited image (e.g., red-haired version).

  • Comparison grid (original + variants).

6. Notes

1. Precision Tips:

   • Use FluxKontextPro for colors/textures.

   • Use FluxKontextMax for poses/structural changes.

2. Performance:

   • Downscale large images with EmptyLatentImage first.

   • Disable PreviewImage for faster processing.

3. Troubleshooting:

   • Unapplied edits: Verify prompt translation.

   • Misaligned grids: Ensure uniform image dimensions.

Example Steps:

1. Load 055.jpg, enter prompt "red hair + standing".

2. FluxKontextPro outputs red hair, FluxKontextMax outputs standing pose.

3. Concatenate all versions with ImageConcatMulti.