Prompting for prompts

Prompting for prompts

The next level up generating AI images is ChatGPT prompting for Midjourney prompts.

By creating a contextual, meta prompt, you can generate expanded, descriptive Midjorney prompts based on simple input, over and over again.

An example for such a prompt is my Selfie Time Capsule, where I generated a ChatGPT meta prompt, telling it to generate a descriptive Midjourney prompt, given a particular subject.

For example, given “Aztec warriors” in ChatGPT would result ChatGPT generating a Midjourney usable prompt:

/imagine prompt: Three Aztec warriors huddled together for a wide-angle selfie picture::5 Wearing feathered headdresses, ornate body armor, and holding obsidian weapons, they stand in a sacred temple surrounded by intricate carvings and offerings. The image has a warm and earthy color palette, with reds, yellows, and greens evoking the vibrant nature of the Mesoamerican culture, showcasing their fierce and proud expressions. The image is photorealistic, 16K, has natural lighting, and is taken with a front-facing phone selfie camera held by one of the warriors in the photo::4. --ar 3:2 --s 1000 --v 5

I told ChatGPT how to start and end the prompt, and for a given subject, to look up the period, the clothing and tools, pick an approriate color palette and facial expressions. That way, I can provide, “WW2 marines”, “Four 1980s Madonna fans”, “three 1950s airplane travelers”, “Genghis Khan”, even “koalas”, “two T-Rex”,… and ChatGPT provides me a Midjourney prompt, which it seems to interpret quite well.

Check out some of the results on this Selfie Time Capsule Instagram account.

Interested in the ChatGPT meta prompt? Find the prompt info on this Selfie Time Capsule landing page, or for the price of a coffee, go straight to Gumroad.

And if you ever wondered what happened with your public selfies on social media, well now you know, it taught AI to generate new selfies…