%20(1).png)
IMAGE & VIDEO GENERATION
I've been working with AI image generation since 2022 — starting with Midjourney v3 and Stable Diffusion, before the interfaces were clean and before the defaults were good. That early period required understanding models at a mechanical level: how prompt weighting worked, how to construct negative prompts, how to coax consistency out of systems that fought it. Since then my toolkit has expanded across every major model — Midjourney, GPT4o, Firefly, Google Gemini Flash Image (Nano Banana), Runway, Kling, and purpose-built fine-tuned models trained on specific visual identities. Four-plus years of daily use means pattern recognition that can't be shortcut — knowing which model handles which brief, when to composite, when to iterate, and when to start over.
PROMPTS WRITTEN FROM EXPERIENCE
Most AI prompts are written in consumer language. The results look consumer-grade. I write mine from a foundation in 3D modeling, photography, graphic design, and digital illustration — which means I'm thinking in camera placement, focal length, light rigs, exposure logic, color temperature, compositional grammar, texture, and visual weight before a single word hits the prompt field. Each discipline adds a layer of precision that generic prompting doesn't reach. The difference shows up in the images.
SYSTEMS AND TOOLS, BUILT IN-HOUSE
Besides generating unique images, I am also able to use this technology to create variations in existing images. I have found this to be extremely valuable in creating site product imagery, which utilize many of the same, repetitive elements. It is also extremely valuable in performance creatives, as I can turn one image from a winning creative into many, each with subtle changes, ready to be used for testing variations.
OPERATIONS
I don't use AI because it's faster — I use it because it extends what I'm capable of. The creative decisions, the art direction, the taste — those stay mine. What AI handles is execution at a scale or speed that would otherwise be impossible. That distinction matters. The designers who will struggle are the ones who handed over their judgment. The ones who thrive will be the ones who kept it and built better tools around it. I'm in the second group.
VIBES
My site runs on Wix, but it doesn't stop there. I extend it with custom code and AI-assisted development to hit a visual spec the default tooling can't reach on its own. The aesthetic — the hierarchy, the layout decisions, the interactive details — required implementation that goes beyond what any template offers out of the box. Wix handles the foundation. I handle everything on top of it.
IMAGE GENERATION
I've been working with AI image generation since 2022 — starting with Midjourney v3 and Stable Diffusion, before the interfaces were clean and before the defaults were good. That early period required understanding models at a mechanical level: how prompt weighting worked, how to construct negative prompts, how to coax consistency out of systems that fought it. Since then my toolkit has expanded across every major model — Midjourney, Flux, Firefly, Ideogram, Runway, Kling, and purpose-built fine-tuned models trained on specific visual identities. Four-plus years of daily use means pattern recognition that can't be shortcut — knowing which model handles which brief, when to composite, when to iterate, and when to start over.