Build a visual baseline. Reuse it forever.

Welcome to Prompt, Tinker, Innovate—my AI playground. Each edition gives you a hands-on experiment that shows how AI can sharpen your thinking, streamline your process, and power up your creative work.

This week's playground: Create one AI-generated visual baseline you can reuse for every future image

Every marketing team has the same AI image story. Someone generates a great image for a blog post. A week later, someone else needs an image for a social post and generates a new one. It looks completely different—different palette, different style, different vibe. Two weeks later, a third image. Three different brands.

The fix is not better prompts. The fix is a baseline.

A visual baseline is one reference image that captures your brand's AI-generated look—the colors, the style, the mood, the way subjects get treated. Once it exists, every future image inherits from it. You stop generating from scratch. You start generating from a foundation.

That's the experiment for this week.

Why this matters

Most marketers I talk to have given up on AI image generation for one reason: it's not consistent. They tried it for a campaign, got five mismatched outputs, and went back to Canva templates or stock photography.

What they're missing is that the consistency problem was a tool limitation in 2024. It isn't one anymore. Both Gemini's Nano Banana and ChatGPT's image generator can now hold a reference image's style across new prompts. That single capability turns AI image generation from "random output machine" into "a system you control."

You only have to do this work once. After that, every image your team generates can hand the baseline to the model and inherit the look. No more "which of these eight images looks the most on-brand?" debates.

Use case spotlight: The brand baseline as a shared asset

Think of the baseline like a logo or a font kit. You don't redesign it for every project—you reuse it. You share it across the team. You version it when the brand evolves.

The teams getting the most from AI image generation right now are the ones treating their baseline as a real brand asset:

• It lives in a shared drive or brand kit
• It's versioned (v1, v2 as the look evolves)
• Anyone on the team can grab it, attach it to a prompt, and generate something on-brand in two minutes
• New hires get pointed to it during onboarding

What used to be "I'll generate something and hope it looks right" becomes "I'll grab the baseline and prompt from it." The output gets better. The work gets faster. The brand stays consistent without anyone having to police it.

Your AI experiment: Build your baseline in 10 minutes

👉 Time to tinker: You'll generate one baseline image in Gemini and/or one in ChatGPT, then test both by spinning up new images that reference the baseline. By the end, you'll have a reusable asset and proof that it actually holds.

Step 1: Write your style anchor

Before you open a tool, write down four things. Keep it short—each one is a phrase, not a paragraph.

• Style: The visual treatment. Examples: editorial photography, flat illustration with thick line work, retro print ad with halftone texture, soft watercolor with grain.
• Color palette: Three to four colors. Use your brand colors or describe them in plain language ("deep navy, warm cream, muted coral, charcoal").
• Mood: One or two words. Confident. Quiet. Optimistic. Grounded.
• Subject treatment: How people, objects, or scenes appear. Examples: "people shown mid-action, no direct eye contact" or "objects on minimal backgrounds with soft shadows."

This four-line description is your style anchor. You'll paste it into every prompt going forward.

Step 2: Generate your baseline in Gemini (Nano Banana) or ChatGPT

Open Gemini and/or ChatGPT. Paste this prompt, select "create image" in the tools section, and fill in your style anchor:

Generate. Look at it. If it's not right, iterate on the style anchor description—don't move on until you'd be happy using this as the reference for everything that follows. This is the foundation. Build it well.

Step 3: Test the baseline by generating three new images

Now prove it works. Open a new chat and upload your baseline image and prompt the next three:

Try three different scenes—a person mid-task, a quiet object on a desk, an abstract concept like "growth" or "focus." If your baseline is strong, all three will feel like they belong together.

If they don't, your style anchor isn't specific enough yet. Tighten the language and regenerate.

💡 Pro tips

• Save the baseline like you'd save a logo. Put both images (Gemini version and ChatGPT version) in your brand kit folder. Name them clearly: brand-visual-baseline-v1-gemini.png and brand-visual-baseline-v1-chatgpt.png. Future you will thank the present you.
• Save the style anchor as text too. The four-line description is the actual brand asset. The images are renderings of it. If a new model comes out next month, you regenerate the baseline from the same text.
• Version it. When your brand evolves, generate a v2. Don't overwrite v1—you'll want to compare them and see how the look has shifted.
• Add typography in Canva or Figma. Image models are getting better at adding words to images (so try it!), but it isn't always reliable. Don't fight them on it. Generate the visual, add headlines as a layer.
• Audit the baseline quarterly. Look at the last 20 images your team generated. Do they still feel on-brand? If not, refresh the baseline. The look should evolve, but deliberately—not by drift.

What did you discover?

The moment your team starts pulling from the same baseline is the moment AI image generation starts feeling like a brand asset instead of a novelty. It's the difference between "we use AI for images sometimes" and "we have a visual system that scales."

I'm curious what your first baseline looks like. Send me a screenshot if you want a second pair of eyes on it.

Until next time—keep tinkering, keep prompting, keep innovating.

📩 Not subscribed yet? Hit the button at the top. More experiments coming soon.