puma velocity nxt campaign system

Context

Design brief: build a PUMA Velocity NXT campaign direction that scales across OOH, social, web, display, and motion, with AI-assisted production for multi-market rollout. The brief split the work into three tasks: creative direction, key visual, automation thinking. I owned the concept, the system, and every tool decision inside that structure.

Insight

Velocity NXT sits between performance running and streetwear. Most sneaker campaigns pick one lane: hero-athlete performance shots, or lifestyle fashion imagery. Either read costs the other audience. The shoe needed a story that was true to both without choosing.

Concept

The city reacts to the wearer's movement. One repeatable idea: everyday urban motion, catching a train, walking a plaza, cutting across a wet court, reads as athletic when the environment responds to it. The shoe leads every frame, closest to camera, sharpest in focus. No studio pose, no staged sport.

Three key visuals, one rule, three intensities of the same reaction:

• Transit Reaction: a closing train door, stopped at the last second. The most urgent beat: motion blur and tunnel light against a sharp, grounded shoe.

• Street Reaction: a walk with coffee, pigeons scattering. The quietest beat: proof the rule holds even when nothing dramatic is happening.

• Court Reaction: a slide across a wet court. The most physical beat: the shoe engaging the surface directly, splash and reflection doing the performance storytelling instead of a pose.

System

The concept became five fixed rules: low-angle, product-first framing; real city scenarios, never staged sets; environmental reaction as the only permitted "action" (train blur, scattered pigeons, puddle splash, light streaks); a neutral city palette with the shoe as the sole colour accent; every visual built to double as a static poster and a motion asset.

The harder problem was production, not the moodboard. Fifteen markets and ten languages will dilute an idea faster than they'll blow a deadline. So before touching any AI tool, I split the system into three layers: a brand-locked core (logo, type, colour tokens, compositional grammar) that stays human-only, permanently; an AI-adaptive layer for hero imagery and environment variants; an automated localization layer for copy. I drew that line myself. It's the actual deliverable of the automation task, not a footnote.

Hero images were AI-generated and upscaled in Magnific (Reimagine, Relight, Background). I directed the prompts, then selected, rejected, and re-directed variants for composition and mood, not just image quality. Localized copy across six languages came from Claude, prompted against PUMA's brand voice, tone, and character limits rather than run through literal translation. I spot-checked every line against the master before it counted as approved. The master key visual, the layout tokens, safe areas, and the compositional grammar were built by hand in Figma. No automation touches those.

Application

That system had to hold from a 4000×2000 OOH board down to an Instagram Story crop, translated into six languages with market-appropriate talent: same shoe, same idea, different city.

Outcome

This was a design exercise, not a live campaign, so there's no media result to report. The proof is whether the system holds under its own rules. It does: one locked master key visual extended into five formats (OOH, web banner, poster, Instagram feed, Instagram story) off shared tokens, with no rebuilding per format. Copy localized into six languages using tone-preserving AI translation instead of direct machine translation. A seven-step production pipeline (shoot, AI polish, master KV, format templates, localization, batch export, human review) designed to support the brief's target of 15 markets and 10 languages without redesigning the idea for each one.