Color Your AI Canvas: ComfyUI_PaletteDirector Guides Image Generation

    
        By vramkickedin    
     | 
    
            June 28, 2026 at 3:36 pm        
    
     | 
    
        2 min read

ComfyUI_PaletteDirector is a set of custom nodes that pairs visual color selection with a trained LoRA to steer image generation in ComfyUI. It builds a palette-card image and adds a structured prompt suffix naming five color roles such as background, main, and accent. The system is intended for use with the provided Palette Director v1.0 LoRA for the Flux.2 Klein model.

SKBv0 that also created ComfyUI_ShowMe developed this project as a hands-on learning experiment in palette conditioning. He built both the custom node pack and the accompanying LoRA, then shared them along with example workflows. The result is less about locking exact hex colors and more about giving the model a stronger directional push than typical color words in prompts.

A palette conditioning experiment for Flux.2 Klein

Key Features

Visual palette composer for easy color selection.
Structured prompt suffix with five palette roles.
Generates palette-card image for LoRA conditioning.
Manual hex color and weight input support.
Positive and negative conditioning text outputs.
Example workflow with Flux.2 Klein included.
Works with the provided Palette Director LoRA.

This tool is for ComfyUI users who want to experiment with palette-based guidance in their image generation pipelines. Hobbyists and curious developers working with Flux.2 Klein can use it to test how a LoRA-driven palette card influences output, though it is not intended for production use. The experiment offers a hands-on way to explore color direction without relying solely on prompt engineering.

Developer notes and known limits

SKBv0 emphasizes that this is an amateur learning project, not a polished professional tool. The palette guidance is not hex-accurate; colors from the prompt can override the palette, and the palette card may occasionally leak into the final image as artifacts. He plans to run additional training iterations over time to address some of the current limitations and improve role consistency.