Everandeverandever
After successfully generating campaign-level imagery of a single product entirely with AI tools, I expanded with this project in scope and complexity: focussing on building the brand and the people around a product.
Left to right: the source portfolio image I had captured and used as input when testing AI models. Figures 2–4: the final shoe with a midsole “glow” to show the active side of the cloud technology.
Why it exists
Coming from a sports background, I've followed the continual development of midsole technology from Air, to Shox, to Ultraboost to Dad shoe.
My first test prompt, in November 2025, was “make a shoe with a cloud midsole using this green shoe as the shape and base reference.”
The shoe was purposely simple in design. Prior testing with models at the time, on other earlier projects, showed logos and text details were a particular weak point.
The fabric upper was prompted to read as a single-piece knit, like Flyknit.
The brand and people around the shoe evolved from that starting point of absurdity, then were developed and presented as realistically as a real brand would be.
The name everandeverandever came from a line in the film Click.
everandeverandevershoe
The design follows modern shoe construction: bonded and laser cut. The shoe is engineered at a modular level to be truly omni-use.
Specification
In use
The athlete clip is demonstration / TVC style. The shoe is designed to behave like other cushioned shoes, absorbing impact and returning in a spring-like feature of modern race shoes. The difference is that these are active clouds, so the shoe is truly multi-purpose: zoned cushioning on a basketball court, then condensing into a denser, storm-like cloud structure for trail running, where stability and traction demands differ from asphalt.
EverCloud Capture Pro-Max-Ultra
The backpack was the link between the shoe, the lab and brand. Exploring its inclusion helped shape the story.
Specification
55, Kenyan-Canadian. Former elite marathon runner, PhD in Particle Physics from UBC. Invented the cloud capture technology after studying high-altitude moisture patterns during training runs in the Rift Valley. Calm under pressure. Speaks slowly, thinks fast.
52, Japanese-Canadian. Industrial design at the RCA, London, plus an MSc in Materials Science. Developed the containment membrane that holds captured cloud under load without collapse — the problem everyone else tried to solve with pressure. Bridges aesthetics and function. Known for asking "but will it feel right?"
41, Barcelona-born. Former footwear designer at a major athletic brand that rhymes with “bike”. Designed the lattice cage that gives the midsole its structure, after eleven prototypes that all deformed the same way. Hands-on maker, always has something unfinished on the bench. The team's momentum — keeps things moving.
24, Chinese-American. Recent SFU graduate, engineering and design. Found the thermal fault in the first capture rig by breaking it, which is how the team learned the device needed a second seal. Eager, clumsy, accidentally creates most of the lab's interesting problems. Aspires to run a marathon in 5 years, had Kofi as his wallpaper growing up.
The lab and team
I built the lab and team around the archetype of an Apple or Dyson: seamless integration between engineering and design, with proprietary technology built in-house at the foundation.
The lab itself was the most time-consuming to design. The wide shot needed plausible function, but also key areas distinct enough to generate scenes within. Splitting design on the left and engineering on the right was intentional: the symbolism of the teams meeting at the centre, but also easier context for the model to follow when prompting.
I gave each member a full backstory to help contain drift when prompting. If Aoki and Elena were in a scene together, Aoki had seniority in the prompt, and the model generated behaviour that better matched.
AI models
Traditional tools
Notes
Athlete clip rebuilt and expanded in September 2026. Main film updated with a few re-run scenes in September 2026.
All generations are native — no upscalers used.