Generative AI Imagery
Image prompting · Prompt frameworks · Figma Weave · Nano Banana · Concept visualisation
I use image models to test campaign looks before money goes into a real shoot.
Outlaw Cap
RaceArt · 2026 · Founder, own brand
AI product shots for RaceArt, my own brand, made in Figma Weave.
Why AITest the look before the shoot
Product photography for the Outlaw Cap launch, made in Figma Weave to try directions before paying for models, cars and locations.
- ProductBlue background, the cap from several angles.
- LifestyleOn vintage car hoods, dashboards, skateboards, people in real alleys.
The prompts put the cap inside each scene and held RaceArt's teal palette, so every image reads as one campaign.
OutcomeA library to launch with
A complete image library for the launch, and a clear brief for the real shoot: I knew which contexts worked before spending on it.
Outlaw Everywhere
RaceArt · 2026 · Founder, own brand
AI campaign for RaceArt, my own brand, 16 prompts in Nano Banana Pro.
IdeaA calm wearer in an escalating world
The world around the cap keeps escalating and the person wearing it stays unbothered. That tension is the whole campaign.
- An igloo an NSX just crashed through
- An arctic snowfield next to a frozen RX-7
- An airport tarmac, an AE86 strapped to a flatbed
- A highway overpass with an Evo IX blasting underneath
- A coal train at 2am
- A tire shop with an S15 on jack stands
MethodHow it was made
Generated in Gemini 3 with Nano Banana Pro. Hard on-camera flash for a paparazzi look; every car JDM, because that's where RaceArt lives. The cap is never shown the same way twice: backwards, pulled low, on a table, clipped to a vest, under a helmet.
Describe a grease smudge on a cheek and you get an image. Describe a mood and you get nothing.
So the prompts stuck to concrete physical details. The launch got a full image library and knew which directions land before booking real models, cars and locations.
After the Meet
RaceArt · 2026 · Founder, own brand
AI campaign for RaceArt, my own brand, a reusable prompt framework.
ConceptWhat happens after the racing stops
An AI photo series about car culture between the action: crews at convenience stores, parking lots and empty streets.
Someone leaning on a yellow car outside a FamilyMart at 2am tells you more about the culture than any staged action shot.
MethodThe prompt framework
Each prompt, 100 to 120 words, pins down:
- Cameraa named camera model
- Placereal chains like 7-Eleven and FamilyMart
- Framing"stolen moment", "candid", to steer the AI away from posed photography
The RaceArt cap turns up the way it would in real life: worn, left on a hood, resting on a trunk. Never posed as product. The framework keeps producing new scenes with the same documentary feel.
Your Plate Just Got Crunchier
Lay's · personal project · 2026 · Personal spec project
Spec ad for Lay's, concept and AI visuals.
InsightThe chip is already a plate
People stack cheese on Lay's, fold salami over them, scoop guac with them. The chip carries everything else and gets eaten with it.
IdeaA bag that becomes a charcuterie board
The behavior already exists, so the visual only has to show it. I made it with AI to build the concept fast and show the thinking: Lay's as the base layer for what people actually eat.
