The problem isn't your AI image tool. The problem is you're not giving it a specific enough job to do. Here's the briefing framework I use before touching any image tool, plus 50 paste-ready prompts from a designer who built visuals for Warner Music and Spotify.
If your AI-generated images consistently look like AI-generated images, there is exactly one thing to check first. Not the tool. Not the model. The prompt.
Every AI image generator is trained on millions of images and, when you give it a vague instruction, it does the most predictable thing it can with that input. It produces the statistical middle of everything it has ever seen matching your description. "A modern brand visual." "A sleek product shot." "A professional headshot background." The model averages every modern brand visual it has ever processed, hands it back to you polished, and you wonder why it looks exactly like everyone else's output.
This is not a technology limitation. It is an input problem. Fix the input, fix the output.
Most prompts are too wide. They describe a category of image rather than a specific one. The simplest test: could your prompt describe ten thousand different images? If yes, the model will produce something that blends all ten thousand together. That blend is what generic AI design looks like.
Designers who have been doing this at a professional level figured out quickly that the tool rewards specificity the same way any creative brief does. A great creative brief does not say "make something that looks good." It names a mood, names references, names what to avoid, and specifies non-negotiables. The output from a great brief is never generic because the brief made it impossible to be.
The system below applies that exact logic to AI image prompting.
Before I open Nano Banana Pro, Midjourney, Sora, or any other image tool, I run a briefing pass in Claude. Claude interviews me, forces me to be specific about things I would otherwise skip over, and hands back a structured brief I can drop directly into any image model.
This extra step takes five to ten minutes and it is the entire reason the output stops looking generic.
Copy the prompt below into Claude. It will ask you six questions, one at a time. Answer each one honestly. Vague answers get pushed back. That pushback is the point. The brief it produces at the end is paste-ready into any AI image tool.
You are my AI design brief writer. I'm about to use an AI image tool to create a visual. My output tends to look generic. Your job is to fix that by writing me a brief specific enough to escape the AI median.
Ask me the following six questions one at a time. Wait for my answer before moving on.
1. What am I creating? One line. Logo, social post, ad, product mockup, illustration,
packaging, or something else.
2. What is it for? The specific use: Instagram carousel, paid ad, pitch deck,
business card, website header.
3. What three emotions should it create when someone looks at it? Real emotions,
not adjectives. "Premium and quietly intimidating," not "professional."
4. Two specific real-world references. Two brands, photographers, designers,
films, or albums whose visual style I want to feel adjacent to. Be opinionated.
5. One anti-reference. One visual style I absolutely do not want to look like.
Name it specifically.
6. The non-negotiables. Specific colors, copy, imagery, dimensions, or
composition rules I cannot change.
Once I answer all six, write me a paste-ready visual brief in this format:
TITLE / USE CASE / EMOTIONAL TARGET: the three emotions plus the specific visual choice that creates each one / COLOR PALETTE: three to five colors with hex codes and the role each plays / TYPOGRAPHY: named typefaces or specific styles / COMPOSITION: where the eye should go, density, hierarchy / LIGHTING AND TEXTURE: the cinematography choices / REFERENCE A: one specific brand or work to borrow from, with what specifically to take / REFERENCE B: a second specific brand or work / ANTI-REFERENCE: what to actively avoid / NON-NEGOTIABLES / THE PROMPT: four to six sentences in second person, opinionated and specific, paste-ready into any AI image tool.
Be ruthlessly specific. If I am vague in any answer, push back. The purpose of this brief is to escape the AI median.
Zero references gives the model the median. Ten references confuses it back toward the median. Two strong, specific references give the model enough constraint to land somewhere specific in the design world without averaging itself back to nothing.
The references do not need to be from your own industry. A beauty brand can reference a film director's visual language. A business coach can reference an architecture studio. The more unexpected the reference, the more distinctive the output.
Amir Mushich, a designer who has built visuals for Warner Music and Spotify, published 50 of his best Nano Banana Pro prompts for free. These are not generic prompts. They are structured, specific, and worth studying as much as using.
Do not just copy them. Read them first. Notice the level of specificity. Notice what he names, what he excludes, and what constraints he builds in. Then use the briefing system above to write your own in that same shape. That is the difference between having 50 prompts and being able to write any prompt you need.
You can find his full library here: Amir Mushich's 50 design prompts on X →
These prompts are primarily built for Nano Banana Pro but most translate directly to Midjourney, Sora, and DALL-E with minimal adjustment.
Any time you produce an image that actually looks like what you wanted, save the prompt that made it. Drop it in a Claude Project or a Notion page. Label it with what it produced and what it was used for.
Organize your saved prompts: social, ad, deck, illustration, product shot, header. Over a few weeks you build a working library pre-tuned to your brand, your tools, and the results that have already worked.
The next time you need a visual, start from the closest working prompt in your library and adjust it for the new job. Speed goes up. Consistency goes up. Generic output goes away.
What happens after a few months of this is that you start to see the pattern in your own work. You develop an instinct for what makes a prompt specific enough to work, and you start spotting generic AI design in other people's content immediately. That instinct is worth more than any individual prompt.
Free Download
Your future self will thank you for saving this.
Every Tuesday, a new free guide on the AI tools, workflows, and frameworks I'm building into real businesses. No fluff. Just what works.