Design + Engineering · 2025 — Present
Ad Creative Engine
AI-Powered Video at Scale
troon_golf
Sponsored
1,247 likes
troon_golf Summer rates across 900+ courses. Book your round today 🏌️♂️
2 hours ago
The Challenge
Troon manages marketing for 900+ courses. Each course needs fresh ad creative — social videos, promotional content, seasonal campaigns. A traditional creative team can produce maybe 10-20 unique videos per week. We needed hundreds. The math didn't work with humans alone.
The Approach
Identified the pattern: most golf course promotional videos follow a similar structure — aerial course footage, key amenity highlights, pricing or offer callouts, and a booking CTA. The variation is in the content, not the format.
Built a pipeline using AI-generated footage (Luma AI) for cinematic course visuals — slow aerial dollies, golden hour shots, clubhouse approaches. No need to send a film crew to every course.
Designed a programmatic video composition system using Remotion. Templates handle typography, animation timing, brand overlays, and dynamic content injection. Swap in a course name, a rate, a hero shot — and a polished 15-second ad renders automatically.
Created multiple creative templates: Course Spotlight for hero content, Play More Golf for promotional offers, seasonal variants for holidays and events. Each template is a React component that accepts data and produces broadcast-quality output.
The Pipeline
The system works in three stages. First, AI generates cinematic footage based on course context — we learned through iteration that slow aerial movements and natural lighting produce the most realistic results. Fast pans and dramatic camera moves break the illusion. Second, the footage feeds into Remotion templates where typography, brand elements, and dynamic content layers compose the final video. Third, the output renders at broadcast quality, ready for social distribution. The entire pipeline is code — every template is a React component, every animation is programmatic, every variable is data-driven.
Designing for Automation
The hardest part wasn't the technology — it was designing templates flexible enough to handle any course while maintaining quality. A template that looks great with a 4-word course name breaks with a 12-word one. A sunset color palette works for a desert course but not a Pacific Northwest one. Every design decision had to account for the full range of inputs. This is a different kind of design problem — you're not designing one output, you're designing the system that produces thousands of outputs.
Outcomes
Video production time reduced from weeks to minutes per course
Scalable system capable of generating unique creative for 900+ courses
Multiple template variants for different campaign types
AI-generated footage replacing expensive on-location shoots