How to prompt ChatGPT to get production-quality images

Whenever I open twitter or instagram, I see AI images that are hyper-realistic as if shot by a world-class photographer.

I assume it's easy and get overconfident thinking I can do the same.

I type a lazy prompt "You are a world class ad creative director. Can you create a ad for the attached product" and I get lazy results.

I try, retry and get frustrated wasting time repeatedly prompting and getting generic images, inconsistent results and not getting what I want!

I call this LILO: Lazy Input, Lazy Output.

I tried to understand how world-class prompters create such production-quality images. Here's the structure.

Every good prompt consists of 3 parts:

• Base Reference
• Idea Detailing
• Camera Detailing

Base Reference:

If you already have an image, character or reference, then half of the problem is already solved. The attached image anchors the model to the attachment.

Idea Detailing:

Answer the following questions in your prompt.

• What is the image type?
  • ex: ad, poster, portrait, headshot, photo-realistic image
• Who is the subject?
  • ex: ferrari car, freckled teen, robotic dancer, redbull can, 3D version of Charmander
• What does the subject do?
  • ex: leaping across a sand dune, peaking through a frosted glass window
• What is the environment?
  • ex: class room, abandoned soviet factory, refreshing water flowing from left to right

Camera Detailing:

• What is the backdrop?
  • ex: gradient shift from orange to red, blurred
• What is the lighting contrast?
  • ex: sharp studio lighting with crisp reflections, focused spotlight from above
• What is the tone?
  • ex: Cinematic, sinister shadows, modern

My structured prompt:

Create a ad for olipop can(attached). The can is centered wit Show water splashing around the can, right to left. Use studio lighting highlight the can but with visible text. Add a black to orange gradient backdrop. Overall look should be refreshing and clean.

Go, Try it on chatgpt and let me know!