I’m experimenting with more detailed prompts for generative AI.
If you haven’t noticed, I use a ton of AI-generated images in Bredemarket blog posts and social media posts. They primarily feature wildebeests, wombats, and iguanas, although sometimes they feature other things.
My prompts for these images are usually fairly short, no more than two sentences.
But when I saw some examples of prompts written by Danie Wylie—yes, the same Danie Wylie who wrote the Facebook post earlier this year at the https://m.facebook.com/story.php?story_fbid=pfbid0nvmhyuLpn3jwMv8K8sbK5EXfS4kcpjfWHicgj4BJhdFLMme87P5fvPSYf9CwjRH7l&id=100001380243595&mibextid=wwXIfr URL—then I realized that I could include a lot more detail in my own image prompts.
If you read Wylie’s Facebook post, or my own subsequent post at the https://bredemarket.com/2025/06/03/when-hivellm-pitches-an-anti-fraud-professional/ URL, then you know exactly what the picture depicts.
Plus some other stuff buried in the details.
By the way, here is my prompt, which Google Gemini (Imagen 4) stored as “Eerie Scene: Sara’s Fake Bills.”
“Draw a realistic picture of a ghost-like woman wearing a t-shirt with the name “Sara.” She is holding out a large stack of dollar bills that is obviously fake because the picture on the bill is a picture of a clown with orange face makeup wearing a blue suit and a red tie. Next to Sara is a dead tree with a beehive hanging from it. Bees buzz around the beehive. A laptop with the word “HiveLLM” on the screen sits on the rocky ground beneath the tree. It is night time, and the full moon casts an eerie glow over the landscape.”
I didn’t get exactly what I wanted—the bills are two-faced—but close enough. And the accident of two-faced bills is a GOOD thing.
How detailed are your picture prompts?
Bredemarket 
1 Comment