(This template is used for Workflow Contest)

What this workflow does

👉 Forces image generation to randomize object/image composition & color

Create a prompt of "cyberpunk city street view" &  the same colors, same one-point perspective, comes up. This forces all images to have more diverse compositions.

Can be used with many other prompts -- like steampunk, architectural, interior or exterior shots. This method is more suitable for non-human-centric images. That's how you achieve a unique and varied visual outcome!

How to use this workflow

👉 3 points of prompt entries: The first section, top & bottom image prompts should be highly geometric shapes with color sets that you want.  The last entry should be the final item desired.

Tips about this workflow

👉 Keep the CN (controlnet) depth strength under 0.5 to give the system freedom to interpret it.

- Keep the CN color strength high for the opposite reason. Feel free to mess with the Image Blend (I've found a simple 50% normal mix to be a bit more predictable with colors)

- Keep the final render step count lower so the system doesn't override the variations

- the original latent should have a max 512 (on one side) to prevent highres oddities

🎥 Video demo link (optional)

👉 [Please add here]