Ads made with generative AI: what works and what does not

What generative AI brings to an ad

Generative AI lets you create images, characters, settings and voices that used to require a huge shoot, expensive visual effects or simply were not possible. That opens the door to ideas a small or mid-sized brand could not consider: a fairy-tale world, an animated character, an invented city. It also speeds up testing: you can see versions of an idea before deciding. What it does not provide on its own is the idea, the judgement to choose among a hundred outputs, or the final finish. Those three things are still the work of a creative team.

Where it works best

It works especially well in three cases. When the idea calls for something impossible or very expensive to shoot: fantasy, absurd humour, exaggerated settings. When you need a character of your own that appears across many pieces. And when you need a lot of content with a consistent look. Two real examples of ours: the Bunker Christmas campaign, with a world inspired by Willy Wonka, AI-generated characters and settings, and two spots built for social; and Canta Coco, a children's song format with its own character, animation and original music, which has passed half a million views on YouTube.

Where it fails

It fails when used as a shortcut to avoid thinking: the result looks like everything else on social media and people spot it as AI straight away, in a bad way. It also fails on telltale details: hands, text inside the image, continuity between shots, faces that change from one moment to the next. And with real products: if people are going to buy something specific, what appears in the ad must be that product, not an invented version. That is why in many projects the best approach is a mix: real footage of the product and the people, and AI for the world around them.

What does not change: idea, direction and post-production

A well-made AI ad resembles a traditional production more than you might think. It starts with a concept, a script is written, the visual world is designed, many tests are generated and discarded, and at the end it is edited, graded, sound-designed and adjusted shot by shot so everything looks like the same film. Post-production is what turns a pile of generated images into a piece with rhythm and finish. There is also a legal side worth checking: the usage rights of each tool, and that nothing looks too much like an existing brand, person or work.

How to approach a campaign with AI

The useful question is not shall we make the ad with AI, but what idea do we want to tell and which tool tells it best. Sometimes the answer is to shoot, sometimes to generate, and almost always a mix. At Bostak Films we use generative AI as one more tool within creative direction, with the same standard of finish as a live-action shoot. If you have an idea that seemed impossible to produce, or want to know whether your campaign would gain anything from AI, tell us about it: from our creative agency in Barcelona we will tell you frankly which part to generate and which to shoot.

FAQ

Got a project in mind?

We tell stories in video end to end — production, filming and post-production.

Keep reading