Microsoft released an ad created with AI, but no one noticed

Three months ago, Microsoft released a commercial for its various Surface devices that focused heavily on AI. But it turns out that AI was playing a much bigger role in the ad than first appeared. According to a new post on the Microsoft Design blog, much of the work on the video was done by AI.

The Visual Designt team had a month to create a commercial and decided to use generative AI to bring their ideas to life faster. They first used AI to write a script, create a storyboard, and pitch the finished idea, which took just a few days instead of the weeks it usually takes.

The next step was to use generative models to create videos. But before they could get the footage they wanted, the Microsoft team had to formulate queries, which they also turned to AI for. They first used natural language to describe specific scenes to the chatbot, and then asked it to convert those descriptions into a fast format suitable for use in generative tools.

"We probably went through thousands of different prompts, chiseling away at the output little by little until we got what we wanted. There’s never really a one-and-done prompt," says Cisco McCarthy, creative director at Visual Desit.

The team also used other creative methods, such as taking photos of compositions that matched their creative direction and sending those images to an AI chatbot, which provided textual cues to help them get closer to the desired result.

As a result, many of the shots that appeared in the commercial were created using artificial intelligence, in particular thanks to large-scale speech models for video generation, such as Hailuo and Kling. However, some scenes still had to be shot manually, in particular those where people type on a keyboard, as even the most advanced models still have difficulty in such situations, despite significant progress in recent months. Instead, shots with quick transitions and limited movements were successfully generated by artificial intelligence.

Using AI has significantly accelerated the filming process for Microsoft. To capture such footage in real life, the team would have had to travel to different locations or recreate complex logistical scenarios.