Blend logos into AI generated images

How-to and samples

Workflow

I created some generative images with prompts that also hides an iconic logo. Can you spot the brand in these images?

Prompt: runner in foreground, running path, landscape, nature, lake surrounded by trees , blue sky, white clouds, modern, drawing, high angle, excellent composition, aesthetic

Prompt: apple tree, red apples, orchard, green meadow, rustic, excellent composition, aesthetic, ultra detailed

For negative prompt, i always used: ‘low quality, lowres, blurry’. Here you can see the difference between a lower and higher value for the illusion strength:

Prompt: fast food restaurant, open space, windows, burgers, fries, high detail, modern

Prompt: fast food restaurant, open space, windows, burgers, fries, high detail, modern

It even works with text logos:

Prompt: japan, skyline, lights, skyscrapers, night, moon, mountain range, high detail

They can even be used for a somewhat darker narrative:

Prompt: sweatshop, big hall, bad lighting, workers, long rows, tech, excellent composition, aesthetic, ultra detailed

Prompt: beach, oil spill, disaster, birds covered in oil, tar, excellent composition, aesthetic, ultra detailed

Technical bits

This is the viral image that started. The workflow was discovered by a person who goes by the moniker MrUgleh

Image by MrUgleh (source)

The novelty here is that it uses the stable diffusion image generator and adds a constraint on top called ControlNet. This can be configured with a ControlNet conditioning scale that acts as the illusion strength. I have used values in the range of 0.6 to 1.5 in the samples above.

Want to make your own? You can do it here. Be sure to have a black and white square image with your logo and be creative with your wordings (you can have a look at what certain words do at this link for inspiration). Don’t expect good results immediately, the above samples were usually the best out of 5 experiments.

Happy designing! (if designing is still the right word…)