Generative AI Creative · Bremen
Twenty years behind the camera, now directing generative imagery. Real products, new worlds, nothing lost in between.
Each row, one project
Packshot → AI scene → detail
A white-background watch packshot turned into a sunlit liquid set. Dial, strap grain and logo kept intact.
ComfyUI · Nano Banana · Image-to-video
Swiss watches shot in studio, rebuilt in darker worlds. Every gear, screw and engraving preserved.
ComfyUI · LoRA · Flux
A ring on white becomes an editorial still life. Stone facets and metal reflections under control.
ComfyUI · Nano Banana Pro
Glitter serum in glass: suspended flakes, refraction and label stay true from packshot to campaign.
ComfyUI · Nano Banana
One bottle, two worlds: a sculpted studio set and hard poolside light.
ComfyUI · Flux · Image-to-video
From ghost mannequin to a model on location. The print has to match, scale for scale.
ComfyUI · Identity LoRA
Leather, zips and buckles: the hardest garment details, worn in an urban night scene.
ComfyUI · Identity LoRA
Three archive pieces, three saturated worlds, one art direction.
ComfyUI · Nano Banana
Sportswear shot in studio, then worn by AI models: dawn run, court, brutalist gym. Plus a running loop.
ComfyUI · Image-to-video
Packshot, ghost mannequin or on-model shot on white. Every angle and detail documented, because the AI can only respect what it can see.
ComfyUI workflows combining local and cloud models, LoRA and structured prompts, designed around the brand's visual rules and the volume to deliver.
Light, perspective, textures, garment drape, anatomy and product fidelity, checked image by image before anything leaves the studio.
I have been a professional photographer since 2007: fashion, e-commerce, luxury, architecture and editorial, including ten years in London.
At QVC UK I initiated a team dedicated to bringing generative AI into photography, and trained photographers and freelancers in the new methods.
Today I design ComfyUI pipelines for high-volume e-commerce production in Bremen, and I build systems for the hard problems of generative imagery: consistent identity, multi-view generation and virtual casting.