AI-generated event posters have become a running joke. Bunting, pastel florals, that same airbrushed “craft fair” look, repeated on every village noticeboard in the country. A guide shared on Hacker News shows the fix is embarrassingly simple: stop accepting the default, and ask ChatGPT what else it can do.
The author’s point is sharp. The posters aren’t bad. They’re fine. The problem is repetition. See that style 20 times and it starts to grate. And since ChatGPT knows hundreds of design traditions, there’s no reason to settle for the one it reaches for first.
Quick Start
What you’ll learn: how to steer ChatGPT away from its default poster template toward a distinctive design style, and how to build a menu of styles you can request by name.
What you need: ChatGPT with image generation, your event details, and about 15 minutes.
Step 1: Give it the details and try to specify a style
The author started with invented event details and an explicit request to avoid the usual look:
“Produce a poster for a spring fayre. 21 April – 11am to 3pm Mill Beach Park, Honeyford Free entry Tombola Cakes and drinks Performance by a samba band and a dhol band. Craft stalls Circus Skills workshop Go for a clean, unfussy, bright layout with a bold striking spring-themed graphic. Avoid pastel/airbrush/oil style art or images of people.”
Why this matters: you’re establishing the content and the negative constraints. But note the result, according to Hacker News: it still came back looking like the default craft-fayre template. Telling the model what not to do isn’t enough on its own. Its “spring fayre” prior is strong.
Step 2: Ask for a completely different aesthetic
The second prompt did the real work:
“Make another one using a completely different design aesthetic of your choice. Treat the current one as a “what not to do”, not that there is anything wrong with it, but we want ours to stand out from other posters that look similar.”
The result was a Bauhaus-style geometric poster. Why this matters: you’re using the first output as a reference point for contrast. That’s far more concrete for the model than abstract adjectives like “clean” or “bright.” The author’s verdict: “It stands out just by being different.”
Step 3: Ask it to name the style
Once you get something you like, you need a way to request it again. The author asked: “What is this design style you’ve used called?”
ChatGPT described it as “Bauhaus / Modernist Poster Design” with geometric minimalism and Swiss Style influence, and offered a shorthand label: “Bauhaus-inspired geometric minimalist poster.” Why this matters: a named style is a reusable prompt. Next time, you skip the trial and error and ask for it directly.
One tip from ChatGPT’s own reply is worth keeping. Most local event posters default to bunting, hand-drawn florals, pastel palettes and craft-fair aesthetics. Anything that looks like “a gallery poster” or “a design museum flyer” wins by contrast.
Step 4: Get a menu of styles
Instead of dredging design vocabulary from memory, ask: “Give me a list of styles to choose from.”
ChatGPT returned a varied menu, grouped by mood. Here’s the list as reported:
- Clean but characterful: Bauhaus / Modernist, Swiss Style, Contemporary Editorial (magazine-style)
- Graphic and illustrative: Risograph Print, Cut Paper / Collage (Matisse-inspired), Botanical Scientific Illustration (modernised)
- Bold and unusual: Brutalist Graphic Design, 90s Rave Flyer / Acid Graphics, Memphis Design (80s postmodern)
- Quietly distinct: Japanese Minimal, Monochrome + Single Accent, Wayfinding / Signage Style
- Playful but clean: Modern Icon System, Stamp / Letterpress Style
Why this matters: you don’t need to be a designer. The model already holds the vocabulary. Your job is to pick a direction and ask for it by name.
Why this matters beyond posters
This is really a lesson about defaults. Every generative model has a “most likely” output for a given request, and that’s exactly the output everyone else gets too. The people who get distinctive results aren’t using better tools. They’re spending one extra prompt to escape the average.
The same trick works for ad creative, social graphics, slide decks and landing pages. Generate, ask for the opposite, name it, then build your own style library.
Next steps
- Run the four-step loop on your next event or promo graphic.
- Save the style names that work in a notes file. That’s your personal design vocabulary.
- Try one of the “left-field” options like risograph or brutalist. Restraint matters, but the contrast is where you win.
- Ask the model for style suggestions in other domains too: email layouts, thumbnails, infographics.
The full write-up on Hacker News includes the poster images at each stage, and they make the point far better than words do.