Turning a still image into a video ad

This piece is now part of a longer guide: UGC ads for performance: the complete guide.

Not every asset justifies a shoot day. Some concepts need to exist quickly enough to be tested, and generative video is genuinely good at that specific job now.

When this is the right tool

• The concept needs testing before it earns production budget.
• The product is static and well photographed already.
• The asset is a variation of something that works, not a new bet.

When it is not

• The product has to be handled, applied or worn.
• The ad implies a real person had a real experience.
• It is the hero asset for a launch.

The workflow

Start from a real still. A product photograph you own, correctly lit. Generated products drift in ways that are obvious at full screen.

Write the motion, do not just prompt the scene. Describe the camera move and the beat, not the vibe. "Slow push in, product centred, background falls out of focus" gets a usable result. "Cinematic product shot" does not.

Generate more than you need. The hit rate is not high. Ten generations for two usable takes is normal, and that is still faster than a setup.

Grade it into the family. The failure mode is an asset that looks like it came from a different account. Apply the same treatment as the rest of the batch.

Cut it against real footage. The strongest use is generative b-roll inside a human-led edit, not a fully generated ad.

The line

No synthetic replicas of real people, no fabricated testimony. If a piece of creative implies a person exists, that person exists, was paid, and signed.

Questions people ask

Will viewers notice? For product motion and camera moves, usually not. For faces and hands, increasingly yes, and the trend is moving against you.