Overview
Every performance video ad started the same way: a heavy After Effects project, hundreds of manually exported images, and hours of copy pasting into every size and language variant. When the app UI updated, I had to rebuild every asset from scratch, the kind of pure repetition that makes you question whether you're designing at all. Using Claude Code, I designed and built "Azar Perf Video Generator," a custom Figma plugin that connects Figma directly to Azar's live component library, and turned a 3 hour job into 15 minutes.
The Constraint
Performance marketing video was pure repetition. Every new script meant rebuilding a 10s cut, a 6s bumper, translated copy, and dozens of size variants, just to test which visual worked in the first 3 seconds. And because the assets were static exports with no live connection to the product, every app UI update meant starting over. A single script could take up to 3 hours to become a full set of deliverables.
Before
After
Where it started
This didn't come out of nowhere. A teammate had built a plugin to automate country/language variations for our ASO assets. Watching that, I realized the same variation logic could solve the most painful part of my workflow, and I extended the idea from static ASO images to motion. Good tools spread: I took a teammate's approach and pushed it into video.
What I Built
The Craft : designing for control, not just speed
Ad formats are fairly fixed once locked, but the team needs to keep testing new creative ideas within that format, and a finished video often needs a small fix later, not a full re generation: swap one photo, reorder two cards, change a name.
So instead of a "generate and done" tool, I designed for control: every card, tag, and photo stays individually editable, and adding a new video format later is a config addition, not a rewrite. What made the process smooth wasn't that everything was automated, it was that I kept the right things adjustable.
Knowing which parts of a creative process to automate and which to leave in human hands is itself a design decision.
The slowest part: localization
Language variation is the single biggest time sink. Every ad needs localized copy, subtitle line breaks, and per market UI text, redone by hand for each market. In the plugin, that collapses to one dropdown: pick a market or a focus country set (Azar UK / DE / FR / IT, Vividi Turkey / Saudi…), and it generates the full localized video set, German, French, Italian, English, in a single click.
Results (After)
3 hours → 15 minutes per full deliverable set, with zero crashes, everything rendered and saved natively in Figma.
Supports multiple video templates (New Way & PROM Perf Mode), auto-generates full profile sets per card, and exports across multiple countries and languages in a few clicks.
4th place, Match Group AI Day APAC.
Adopted by the team, still used every time the app UI updates.
My role shifted from producing individual videos to designing and maintaining the system that produces them, deciding what to automate, what still needs a human eye, and where the workflow can go next. I don't just use AI tools; I build them, treating automation itself as a creative and strategic decision, not just a productivity shortcut.
