How To Use

1. Upload a Character Image – Start by uploading an image of the character whose outfit or skin you want to modify.
2. Describe Your Changes – Enter a description of what you’d like to see changed. This could be details about outfit elements, color schemes, textures, or material variations.
3. Define the Style – Provide an overall description of the aesthetic you want to achieve (e.g., concept art, painterly, 3D rendering).
4. Select a Model – Choose the AI model you want to use for generating the variations.
5. Generate Variations – Run the workflow to create different character variations while maintaining the original character’s pose and structure.‍
6. Save Your Output – The results will be saved to a board where you can review and refine them.

How It Works

1. Image Processing – The uploaded character image is processed to extract important details, including edges and depth maps, using ControlNet models such as Soft Edge Detection and Depth Map Control.
2. Text Prompt Integration – User-defined inputs, such as desired style and changes, are combined into a structured prompt that directs the model’s generation.
3. Noise and Latents Generation – A random seed is applied to introduce controlled variations, and an AI model (such as SDXL) generates an initial latent noise representation.
4. Conditioning and Refinement – The AI model refines the generation using both positive and negative prompts. Negative prompts filter out unwanted styles (e.g., blurry or flat images), while positive prompts guide the AI toward the desired outcome.
5. Denoising and Control Layers – Control Layers ensure that the generated variations maintain the structural integrity of the original character while applying the requested changes.‍
6. Output to Board – The final processed image is rendered and saved to a project board where it can be further reviewed, compared, and iterated upon.