AI in video: where it delivers, and where it falls short

With AI in video today you can previsualize an idea, build large-scale scenes, explore art variants, adapt framing to other formats and integrate what is generated with real footage, CGI and VFX. It still requires more control in complex physical interactions, extended continuity, text within the image and products that have to look exact.

The most useful question is no longer whether artificial intelligence can make video. It can. The question is in which parts of a production it provides a real advantage and in which it introduces more complexity than it solves. That assessment depends on the type of piece, the level of precision and the brand context.

Where AI contributes most in a video

• Previsualization: testing ideas, cameras, styles and scenes before closing the production.

• Visual scale: building transformations, crowds, worlds or situations that are hard to film, like the rain and the exteriors of Días de lluvia at Alto Palermo.

• Exploration: comparing art and composition variants more quickly.

• Adaptation: extending backgrounds, rebuilding framing and preparing versions for different aspect ratios.

• Hybrid production: integrating with real footage, CGI, motion graphics and VFX.

In Días de lluvia, a traditional production with multiple characters, exteriors, rain and locations made it hard to be ready on time. Combining AI with corporate material and real assets made it possible to resolve characters, exteriors and weather without splitting creative work and production into long stages.

Where AI in video still requires more control

Complex physical interactions, extended continuity, certain camera movements, text within the image and the exact representation of some products can require more work. The level of demand changes when a piece represents a brand and not just a visual idea.

That does not mean those scenes are ruled out. It means they call for more review, more post-production or a different technique for that part of the piece.

When it makes sense to film, animate or design instead of generating

Some scenes are handled better by filming. Others call for design, animation or traditional compositing. AI is most valuable when it adds options, not when it forces the use of a technique even though there is a simpler or more authentic one. How the technique for each shot is decided is covered in Shoot, AI or both.

How the criteria change depending on where the piece is seen

A scene meant for a reel can tolerate a different kind of detail than a commercial, a large-format screen or a product piece. Before choosing the production workflow, it is worth understanding where it will be seen, how long it lasts and which elements have to be exact. On large-format screens, those criteria change even more: we look at this in DOOH and screens.

How to assess a project by its requirements

Technical capabilities change month by month, so any closed list quickly becomes outdated. The most solid approach is to assess the project by its requirements:

• Consistency: what has to look the same in every scene.

• Realism: how much the piece has to look like live-action footage.

• Length and number of shots: how much material has to be sustained.

• Formats: on which formats it will be seen and in what proportions.

• Product: what has to be exact, with no margin for error.

• Approval times: how many review stages there are on the brand side.

With that list, each part of the piece finds its technique. How those decisions are then organized is covered in From brief to final video, and the full picture in AI video production for brands.

A good production does not use AI because it can. It uses it when it improves the idea, the process or the result.

Frequently asked questions

What can be done with AI in a video?

Previsualizing ideas before producing, building large-scale scenes or scenes that are hard to film, exploring art and composition variants, extending backgrounds and preparing versions for different aspect ratios, and integrating scenes with real footage, CGI, motion graphics and VFX.

What is better not handled with AI in a video?

Complex physical interactions, extended continuity, some camera movements, text within the image and products that have to look exact can require more work. Sometimes filming, animation or design handles the scene better.

How do I decide whether my project should be done with AI?

By assessing it by its requirements: how much consistency it needs, what level of realism, how long it lasts, how many shots it has, in which formats it will be seen, which products have to be exact and how much time there is for approval. With that list, the technique for each part is chosen.

This is how we work on campaigns and commercials at FeedGuru.