AI Training Library · Module 04

Subscriber-only resource.

This module is reserved for newsletter subscribers. If you've already signed up, check your inbox for your personalized unlock link — it works across every module in the library.

Haven't subscribed yet? Free, no credit card, instant access.

Get your unlock link
AI Library · Module 04

One image.
Three brains.

The same brief through Claude, ChatGPT, and Gemini gives you three different images. Here's how to write the brief — and the trick: when stuck, let AI write the prompt for you.
What this module covers

Stop fighting one image tool. Start picking the right one.

Most people pick one AI image generator, get a mediocre result, and stop. The pros run the same brief through multiple tools and pick the best output. Below is the workflow, the meta-prompt that writes your image prompt for you, and the reverse-prompt trick that turns any image you admire into a reusable template.

The workflow One image. Three brains.
  • Ask Claude for the prompt. Claude doesn't generate images — but it writes excellent prompts for the tools that do.
  • Paste that prompt into ChatGPT and Gemini. Compare results. The "best" tool changes depending on the subject.
  • Use the reverse-prompt trick. When you see an image you love, ask AI to write the prompt that would create it.
Step 01 Ask Claude to write your image prompt
I need to generate an image for my business. Write me a detailed image prompt I can paste into ChatGPT, Gemini, or Midjourney. Here's what I need the image for: - Purpose: [where it's going — website hero, social post, ad, email header, etc.] - Subject: [what's in it — a product, a person, a scene, an abstract concept] - Mood: [warm, professional, gritty, playful, calm, urgent] - Style: [photograph, illustration, 3D render, hand-drawn, minimal, vintage] - Audience: [who's seeing it — my customers are roughly...] Write the prompt in one paragraph. Be specific about lighting, composition, color palette, and what should NOT be in the image. Don't explain — just give me the prompt I can copy.
Why it worksClaude knows its limits. It can't make the image, so it writes a prompt good enough that whichever tool you paste it into produces something usable on the first try.
Step 02 When none of the three worked
None of these worked. Here's what came back: [describe the results briefly]. Rewrite the prompt to fix [the specific problem — too generic, wrong mood, weird hands, the people look fake, etc.].
Why it worksEach tool has a house style — ChatGPT goes polished and literal, Gemini goes editorial and clean, Midjourney goes dramatic. Same words, different images. You stop fighting one tool and start picking the right one.
Step 03 — The trick Reverse-prompt: make AI describe what you like
Look at this image. Write me a detailed prompt that would generate something with the same style, mood, lighting, and composition — but with [my subject] instead. I want the prompt as one paragraph I can paste into an image generator. Don't explain it, just give me the prompt.
Why it worksDescribing visual style in words is hard. AI is better at it than you are. This turns any image you like — a competitor's site, a stock photo, an Instagram post — into a reusable template for your brand.
Bonus Reusable image brief template
A [photograph / illustration / 3D render] of [subject], [doing what / positioned how]. The mood is [mood words]. Lighting is [natural / soft / dramatic / golden hour / overcast]. Color palette: [colors]. Composition: [close-up / wide shot / centered / off-center]. Style reference: [a known style — "editorial magazine photo," "1970s film still," "modern minimalist illustration"]. Do not include [things to avoid — text, logos, people's faces, clutter].
Why it worksForces you to think in the categories image tools actually care about. Skip any of these and the tool fills the blank with its default — which is usually not what you wanted.
Reference Tool-by-tool strengths

Each tool has a default aesthetic. When you have a feel for which tool fits which job, you save iterations.

  • ChatGPT (DALL-E 3): Polished, literal, follows instructions closely. Best for: product shots, clean composites, anything where you need the AI to do exactly what you said.
  • Gemini (Imagen): Editorial, clean, slightly stylized. Best for: lifestyle imagery, modern minimal aesthetics, blog headers.
  • Midjourney: Dramatic, atmospheric, sometimes too stylized. Best for: hero images, artwork, anything where mood matters more than accuracy.
  • Ideogram: The only one that handles text well. Best for: posters, signage, anything with words baked into the image.

When in doubt, run the same prompt through two and compare. Takes 90 seconds.

Don't do this Common image-prompt mistakes
  • Asking for "a professional image of a small business." Means nothing. Specify subject, setting, mood.
  • Trying to fix bad output by adding words. If it's wrong, rewrite the prompt — don't pile adjectives on a broken brief.
  • Forgetting the "what NOT to include" line. AI will add text, watermarks, weird hands, extra people if you don't explicitly exclude them.
  • Generating images with text inside. Most tools garble text. Use Ideogram, or generate the image without text and add it later in Canva.
  • Stopping at the first image. Run the prompt 2-3 times. The variance is the feature.