Spec Work · 2026

Crocs: Brief to Campaign

Two complete :15 spec campaigns, comprising fourteen final frames and two finished films, produced solo through an AI-assisted production pipeline under human art direction. Stills and motion generated via API, beat-cut in code with Remotion.

Spec work: not commissioned by or affiliated with Crocs, Inc.

2 :15 Campaign Films

14 Final Frames

6 Stage Pipeline

The Brief

Could one person take a brand brief to finished campaigns?

The test: study a real brand’s tone, product line, and visual language, then deliver two distinct, finished social campaigns using current-generation image and video models, with every creative decision made by a human art director and every frame measured against a single bar: it has to pass as real, or it doesn’t ship.

The hard part of AI production isn't generating images. It's consistency: the same camera, the same product colorways, the same character, frame after frame, until nine separate generations read as one campaign. That's a discipline problem, not a prompting problem, and it's what this project was built to prove.

Nano Banana Pro Seedance 2.0 fal.ai Remotion

Project A

“Every you. Every day.”

One fixed top-down camera. Feet in Crocs, dead center, every frame. Everything else changes. Nine ordinary moments from one life, beat-cut at 124 BPM into a :15 loop. The discipline of the locked composition is what makes nine generations read as one campaign.

Frame 1: sprinkler — The composition lock: every other frame anchors to this camera, scale, and stance.

Frame 2: poolside — Colorway + charm fidelity held against product refs; loaded charms QA’d at zoom.

Frame 3: dog walk — The scene was refined through art direction, with the dog cropped to a tail at the frame’s edge.

Frame 4: grocery aisle — Set dressing kept brand-safe by removing packaging branding in a focused edit pass.

Frame 5: coffee shop — Foreground cup re-staged for physical logic; focus held on product.

Frame 6: beach — Texture retention: sand grain, dusted clogs, no render smoothing.

Frame 7: game night — Character-style charms kept generic; seating verified per hole.

Frame 8: kitchen — Stance lock defended against a graphic ground that kept pulling the pose.

Frame 9: picnic — The mismatched pair is worn on purpose and styled into the campaign.

:15 Film: Project A

Project B

Fashion Swap

One character, one fixed frame, five looks. The Crocs change with every outfit. Hard cuts land mid-gesture; the styling coordination is the content. Character consistency held across every still and clip from a single locked face reference.

Six-panel character sheet: identity locked once and carried through 10 stills and 5 clips.

:15 Film: Project B

Process

Six stages, every select human-approved

Brief

Brand study from the live platform; tone pillars and a hard avoid-list set before any generation.

Refs

A product reference library built from catalog photography, with colorways checked at the pixel level and geometry held consistent.

Lock

One hero frame iterated to approval, then locked as the composition reference for every frame after it.

Stills

Generated against the locks, zoom-QA’d per select; anything that doesn’t pass as real gets cut.

Motion

Micro-motion keeps subjects still while their environments move. Identity refs ride along on every clip.

Edit

Cut grids built in code from each track’s measured BPM; one global grade across every select.

Art Direction Decisions

The calls that held it together

Composition Lock

One hero frame became law for all nine. The approved frame-one image was attached to every subsequent generation as a composition reference. Camera angle, product scale, and stance had to match it exactly. When a graphic ground kept pulling the pose, the lock won, not the generation.

Character Canon

Identity locked once, then defended. Project B's character came from a single canonical face reference shot on mid-gray seamless because white backdrops bleed into final lighting and push skin toward plastic. Any identity marker that drifted between looks was cut from canon entirely rather than patched per frame.

Re-render vs. Edit

Two different tools for two different fixes. Image editing models stay close to their base image. They reliably remove elements and clean surfaces but resist repositioning objects. Strategy: re-render for layout changes, surgical edit for local fixes. Every select zoom-QA'd at full resolution; feed-size review misses seating and mount errors.

Hard Cuts Only

Generative transitions were tested and rejected. AI flash-swaps and morph transitions both read as AI: one felt wrong on the beat, and the other hybridized mid-frame. Every cut in both films is a straight cut in the Remotion edit, with Project B's cuts landing mid-gesture.

Motion Fidelity

The canonical reference rides along on every clip. First motion pass showed face drift and AI skin: image-to-video only sees the plate, so identity drifts. The fix was reference-to-video at 1080p with the character canon attached to every generation. The lesson: upscalers sharpen what's there; they don't fix it.

What This Proves

Consistency is the product

Anyone can generate a good-looking AI image. The difference between a demo and a campaign is whether frame seven matches frame one, with the same camera, the same colorway, and the same charms in the same holes. That takes reference discipline, a requirement that every frame pass as real, and an art director willing to throw out good frames that don't match.

This project demonstrates a repeatable brief-to-campaign pipeline: brand study, reference locks, generation under constraint, zoom-level QA, and a code-driven edit, with a full generation log behind every select. The pipeline produced two campaigns for one brand. The process generalizes to any brand.

Back to portfolio