If you've been following my earlier posts, you'll notice that I like using generated images as visuals. However, this process often yields some unexpected byproducts. For instance, while generating a thumbnail for a YouTube video, I came across several "random" results that had little to do with the original prompt.

Prompt: “Create a YouTube video thumbnail for a project called '[project name]'. The design should be simple and professional, incorporating the project's name prominently.”

Some of the results by leonardo.ai model DreamShaper v7 :

After seeing these unintended results, I decided to challenge various text-to-image models to illustrate the concept of 'unintentionality' from their own "perspectives". Here’s what they came up with.

Prompt : Create a visually stunning and thought-provoking image that represents 'unintentional'.The image should capture the essence, depth, and innovation associated with 'unintentional', all while being accessible and appealing to a general audience.

MODEL: DreamShaper_v7

MODEL: SDX_09

MODEL: RPG_40

I also gave the same prompt  to DALL·E 3 to see if the quality was notably better. The results? Well, I'll let you be the judge.

Prompt : Create a visually stunning and thought-provoking image that represents 'unintentional'.The image should capture the essence, depth, and innovation associated with 'unintentional', all while being accessible and appealing to a general audience.

You can use DALL·E 3 (coming soon on ChatGPT as well) preview on bing image creator.  As a side note, access to previously generated images is currently has a limit on Bing, unlike with Leonardo. So, if you keep generating without saving, you may need to dig through your browser history to access those images again..

Have you also encountered unintended results with AI-generated imagery? I'd like to hear about them.

I m not affliated with any of the platforms mentioned above. 

And if you like generative AI content, here is the instagram account link where I post daily