A 10-Second Social Teaser From One Photo (Luma)
No camera, no crew, no b-roll library. One product photo plus Luma's image-to-video became the teaser that outperformed everything I shot that month.
The teaser that beat all my real footage last month started as a single product photo. Image-to-video in Luma is the least intimidating entry into AI video, and for social-length clips it produces results that read as intentional rather than generated.
The workflow: pick a photo with depth (foreground product, mid-ground surface, background shadow or gradient; flat photos animate into flat videos), upload it, and prompt for camera motion only at first. 'Slow push in, subtle parallax' turns a still into something alive. That single move is 80 percent of what makes a photo feel like footage.
Keep the ambition small and the clip wins. Ten seconds, one move, maybe a subtle element drift like steam or light shift. The generations that fail are the ones asking Luma to invent actions: hands appear, products rotate themselves, and the uncanny valley arrives on schedule. Camera motion plus ambience stays believable almost every take.
My quality loop is cheap: three to five generations per clip, pick the one where the motion stays consistent through the full duration, and reject anything where the background melts. Then standard social treatment, captions, a hook line in the first second, sound that carries half the impact.
The honest framing: this does not replace shooting. It replaces the gap between shoots. Launch weeks, A/B thumbnail tests, 'we should post something today' days. One good product photo and twenty minutes is now a publishable teaser, and the boring photo I almost did not post became my best performer.