Yesterday I walked into the bagel shop around the corner.
I ordered: “Cream cheese on everything.”
And I realized… that’s exactly how many clients sound today:
“AI on everything.”
Haha, it’s funny — but also a little too real.
Because right now we live in a world where AI apps like Veo, Runway, and even that Nano Banana clip have turned into creative candy. From static images to full-blown short films, music tracks, and animations, AI can generate content so good you want to adopt it instantly. Yesterday, if possible.
But here’s the catch: slapping AI on everything isn’t strategy. It’s toppings without the bagel.
The real work is turning these innovations into something that fits your brand. You need to be agile enough to spot early what captures people’s attention, see how other brands are experimenting, and then build a content strategy around it.
Add your twist?
Even better.
Wrap it in your own story?
Now we’re talking. That’s when it becomes more than AI playtime — it becomes a growth.
But let’s be clear:
AI visuals aren’t the finished product. They’re raw materials. Like a photo shoot for your designer to create assets for campaigns. AI doesn’t replace design; it fuels it.
And this is where most brands fall into the trap. Funny or viral visuals might light up TikTok, but they rarely carry the brand DNA — your colors, your typography, your voice, your story. If you drop them as-is into your content calendar, you’re not building equity. You’re diluting it.
A brand isn’t built on randomness. It’s built on consistency. Every post is one brick in a larger narrative. And without that structure, you’re just… posting.
Design on Design
So… what do you actually do to move from random fancy visuals to branded posts?
Here’s where the concept of design on design comes in.
AI generates the first layer — the raw design. But your brand adds the second layer — the real design. That’s when a visual stops being “some cool thing AI spat out” and starts being yours.
Here’s how:
- Start with your brand kit. Collect your logos, icons, and any visual elements you use repeatedly.
- Add your fonts. Use the typefaces that match your company’s voice.
- Layer in backgrounds, masks, and shapes. Keep your Instagram feed or ad sets visually consistent.
- Prompt smart. Mention brand colors, mood, and theme when generating AI visuals.
- Skip the text. Generate visuals clean and add text later with your brand’s fonts.
That’s design on design: AI provides raw creative clay. You sculpt it into a branded asset.
But What About Sizes?
Here’s the next pitfall: you finally got the perfect visual… but it doesn’t fit.
- Facebook wants horizontal.
- TikTok needs vertical.
- Instagram demands square.
- YouTube asks for thumbnails.
If you just crop or stretch, you lose quality and break the brand story. That’s why you need to think in terms of variations.
Variations = taking one master visual and generating multiple, properly framed versions of it — each adapted for its platform. Think of it like cutting a movie trailer versus a TikTok short: same story, different cut.
How ChatGPT & Gemini Handle Variations
When it comes to generating variations, models like ChatGPT (with GPT-image) and Gemini don’t just resize visuals — they work by re-generating the image under new constraints. Instead of cropping or stretching, they use the original prompt (and sometimes the seed image) as a baseline, then apply transformations for different aspect ratios or dimensions.
For example, if your master visual is square (1080×1080) and you need a vertical story (1080×1920), the model doesn’t just stretch pixels — it expands or reframes the scene, filling in missing details to keep composition balanced.
This process is often called outpainting or recomposition. The result: multiple versions of the same creative asset, each consistent in mood, style, and brand feel, but adapted for the technical requirements of Facebook, TikTok, YouTube, or Instagram. In practice, this means you can maintain brand storytelling seamlessly across platforms, without compromising on quality or identity.
How to Prompt for Variations (Step-by-Step)
- Start with the Master Visual Prompt
“Create a cinematic illustration of a coffee shop, warm lighting, cozy atmosphere, brand color palette #123456 and #654321, square format (1080×1080).” - Request Variations with Dimensions
“Now generate variations of the same scene in 1080×1920 (vertical), 1920×1080 (horizontal), and 1280×720 (YouTube thumbnail). Keep colors, mood, and style identical.” - Control Composition
“Ensure the coffee cup remains centered in all versions, even when adapting to different aspect ratios.” - Use Seeds for Consistency
“Use the same seed as the original image to generate all variations.” - Add ‘No Text’ Instruction
“Do not include text overlays.” Add text later in your brand fonts. - Iterate
If an object is cut off in a version, re-prompt: “Reframe so the coffee cup and table are fully visible in vertical orientation.”
That’s the mindset shift: AI isn’t your final design. It’s your raw material. You layer your brand on top, create variations for each channel, and suddenly you’re not just posting AI candy. You’re building a branded content engine.
Because in the end, AI isn’t your brand. You are.
Was this article useful?
Theodore has 20 years of experience running successful and profitable software products. In his free time, he coaches and consults startups. His career includes managerial posts for companies in the UK and abroad, and he has significant skills in intrapreneurship and entrepreneurship.
