AI is part of the job now. These are explorations, art-directing image and video generation to see how it works and how it fits into a real design process. I started with Firefly and Nano Banana, and expanded the projects as I explored other tools like Flora.
I treated the project like a real product launch. I designed the logo by hand, refined it with AI, and built the brand identity before creating the product. After designing the flavour system and can mockups, I integrated them into the campaign visuals. The photoshoot was the final step.
The focus was art direction rather than image generation. I defined the visual language with references inspired by Y2K, Harajuku, summer beaches, 0.5 fisheye photography, and saturated colours. ChatGPT helped me break those references into detailed styling, framing, and pose notes that I used to guide NanoBanana. I generated the models first, integrated the cans afterwards, and iterated until the campaign felt like a real summer launch.
Two key visuals adapted with Flora into four formats each, for different digital ad placements.
A · Too hot? Frost it.
B · Choose your chill
Four UGC-style product videos, built with a very simple setup in Flora. One node holds the product image and connects to a video node. The prompt carries the product references, the shots and the script. I wrote the prompt myself, refined it with Claude, and set the model and the duration.
Swipe →
I explored different textures through close-up portraits, dreamy lighting, holographic skin, and prismatic reflections. Alongside the portraits, I created a clean product still life to build a cohesive visual direction.
The result feels soft, polished, and contemporary, with the lighting doing most of the storytelling. It could easily evolve into the visual language of a skincare campaign.
Social content made with Flora from the Skin7 visuals. Five stories and three feed posts.
Stories · 9:16
Swipe →
Feed · 4:5
Same Flora setup as FROST. The product image goes into a video node, with the references, the shots and the script in the prompt.
The goal was to create a clean, editorial bike campaign inspired by vintage 90s catalogues. I generated the rider and the bike separately before bringing them together into the final visuals.
ChatGPT helped me translate reference images into detailed pose descriptions that guided NanoBanana. I also applied the grain and vintage colour grading directly during generation. The final layout was designed by hand. Composition and typography gave the campaign its finished look.
A six-frame Instagram story sequence made with Flora from the City Frame visuals. One short line per frame.
Swipe →