🤖 AI Summary
Coca‑Cola this week released a second AI‑generated holiday ad — a remake of its iconic 1995 "Holidays are Coming" truck spot — but viewers quickly spotted glitchy inconsistencies: trucks that change shape, wheels that appear or disappear between shots, and a worrying moment that looks like a truck heading toward a crowd. Behind the scenes the brand says five AI specialists stitched together roughly 70,000 video clips in 30 days using tools such as OpenAI’s Sora, Google’s Veo 3, and Luma AI, with Silverside AI and ad agency Pereira O’Dell leading production. The company argues it’s pioneering the use of generative tools rather than waiting for perfection; early testing shows strong attention and brand recall, though some viewers report increased distrust.
The ad exposes a core technical limitation of many generative video systems: poor temporal consistency or “temporal drift.” Many pipelines generate frames or short clips independently and lack a robust memory of prior scenes, so characters and objects can morph across cuts—an obvious giveaway of inauthenticity. That shortcoming has practical implications for marketers and the wider AI/ML community: while AI can dramatically speed production and cut costs, unresolved continuity artifacts risk damaging brand trust, prompting consumer aversion to AI ads and raising questions about quality assurance, safety (e.g., apparent collisions), and the need for models that maintain object and scene coherence across time.
Loading comments...
login to comment
loading comments...
no comments yet