One model, any scene

Project: Self-directed · fictional brand NORTHLINE

Year: 2026

Tools: FLORA, GPT Image, Nano Banana, GPT-5.4

Scope: Brand identity and a full campaign


NORTHLINE is a fictional brand. All imagery was generated with AI as a self-directed practice project.

The challenge

A brand falls apart when every image looks like a different brand, and AI makes that easy to do. I wanted to set the rules once (one model, one outfit, one palette, no stray logos) and have every asset follow them.

The approach

I gave each tool one job. A text step acted as creative director, GPT Image handled anything with letters, and Nano Banana built the photo scenes from one model reference. The rules went into every prompt.

The results

One approved model image fed a whole campaign that reads as one brand. The consistency test held on the face and garments, with small drift in the details.

The process

1. Logo. A bold wordmark with a chevron peak inside the O, built to anchor every asset.

2. Brand sheet. Lockups, palette and labels, checked letter by letter.

3. Campaign idea. A text step wrote the name, tagline and scene prompts. I picked the best.

4. Six scenes. City to trail, same woman, same outfit.

5. Formats. Ads, a five-slide carousel, story ads and phone mockups.

6. Consistency test. Same look, new settings. The face and garments held.

What didn't work

Early brand sheets added photos I hadn't asked for and warped the logo. One character tool kept erroring, so I worked from the approved image instead.

Every detail checked

I reviewed every word and every garment detail by hand.

What I learned

Consistency comes from structure: one locked reference image, the same rules in every prompt, and the right model for each task.

Next: Spyder

Rebuilding a heritage race brand