TH

THE CHALLENGE

​

A gubernatorial candidate in the Northeast needed supplemental ad creative fast. The production timeline and budget didn't support a traditional shoot for every spot. We saw an opening: use Google's Veo 3 generative video model to produce a full slate of broadcast-ready political ads — not as an experiment, as finished campaign creative running in a live race.

​

There was no playbook for this. Nobody had done it at this level in a political environment. We were going to have to build the process while building the product, under a real deadline, for a client with no tolerance for creative that looked unfinished.

​

THE APPROACH

​

The first two weeks weren't about making ads. They were about learning how to make ads with this tool — and that distinction matters.

Veo 3 is powerful, but it requires a specific kind of creative intelligence to produce consistent, broadcast-quality output. The problems surfaced fast: characters looked different from shot to shot without established visual references, audio pronunciation was unpredictable without phonetic guidance built into the prompt, accents appeared randomly in voiceover, and lighting shifted between generations in ways that broke visual continuity across a spot.

​

We solved each problem systematically. Character references were locked first — approved visual profiles that every subsequent prompt could anchor to, ensuring the same face, same wardrobe, same presence carried across all eight spots. We built a phonetic prompting protocol so the audio engine pronounced candidate names, place names, and policy terms correctly every time. We developed lighting and environment descriptors that stabilized the visual tone across generations. Every solution became a documented step in a production workflow.

​

By the time we were generating final assets, the process was repeatable. Not just for this campaign — for any campaign. Anyone on our team can now run an AI video production from brief to delivery using the playbook we built here.

​

The spots worked because we treated Veo 3 like a production tool that required craft, not a shortcut that replaced it. Every problem we hit — the inconsistent characters, the pronunciation errors, the lighting drift — had a creative solution. Finding those solutions took time and discipline. But the discipline is what separated finished campaign creative from the kind of AI content that looks generated.

The more important outcome is the playbook. One campaign's worth of problem-solving became a permanent capability for the team. The next AI video production starts weeks ahead of where this one did — and that's the real value of doing it right the first time.

​

​

​

​

​

​

​

​

​

​