AI vs CGI what each does and when to use which

CGI stands for computer-generated imagery. For a product, it means a 3D artist builds the object in the computer to its real dimensions, from CAD files, which are the manufacturer's design files, or from a sample or photographs. The artist gives each surface its material: glass, brushed metal, printed card. Lights and a camera are placed, as they would be in a photo studio, and the computer calculates the image. That calculation is called rendering.

Because the product exists as a model, it can be turned, opened, taken apart or filmed from any angle, and it is the same object every time. This is what Indian marketing teams usually mean by a CGI ad: a hyper-real product film made without a shoot. Our page on 3D product animation explains how such a film is made.

Generative AI makes an image or a video clip from a description. The person directing it writes the description, called a prompt, supplies reference images, then chooses and corrects the results. Nothing is modelled. The tool has learned from a very large number of pictures what things tend to look like, and it produces a new picture that fits the description.

This makes it quick at things that are slow to build by hand: a city street in winter, a crowd, a period room, a painted style. It also means the tool is estimating. It does not know your product's exact dimensions or the wording on the label unless those are supplied, and then checked in every shot.

Each method is weak where the other is strong. The choice depends on what has to stay fixed in your film.

AI changes details between shots. A bottle cap gains a ridge. The logo is spelled almost correctly. The pack is a little taller in the second scene than in the first. A viewer may not be able to name the fault. A brand manager will, and so will the legal team. Directing AI well is mostly the work of catching and correcting these changes.

CGI goes wrong in a different way. It takes longer to set up, so a late brief leaves little room. A model built in a hurry looks artificial. A fully built world, with streets, weather and people, is far more work in CGI than a product standing on a surface.

VFX means visual effects: changing or adding to footage that a camera has filmed. Removing a wire, replacing a sky and placing a CGI product into a real street are all VFX. CGI is often one of the things used to do it. AI is now used too, for backgrounds and clean-up.

To tell the three apart: CGI is a way of making a picture from a 3D model. Generative AI is a way of making a picture from a description. VFX is the job of combining made pictures with filmed ones. A film with no camera footage at all can be full CGI, full AI, or a mix of the two.

A photograph needs the finished product, a studio and a photographer in the same place. A 3D render needs the design files or a sample to model from. Photography suits a product that already exists and will not change, and it suits food, clothes worn by a person and anything handmade.

Rendering suits a product that is not manufactured yet, comes in many colours or sizes, or needs views a camera cannot take, such as a cutaway or parts moving apart. Once the model exists, a new colour or angle is a change to the file. Which costs less depends on how many images you need and how often the range changes. See 3D product rendering.

For anything that will be manufactured, we model the product in 3D first. That fixes the shape, the label and the colour. AI is then used for what it is good at: the place the product sits in, the light, the atmosphere, and trying ideas quickly before anything is final. The modelled product is placed into those scenes, so it is the same object in every shot.

Our line for this is "Not just generated. Directed." When a campaign is about a world and a feeling, and the product is seen only briefly, an AI brand film can be the right method on its own.

Three things settle most briefs: what must look exactly the same in every shot, how much of the film is the product and how much is the world around it, and how many versions you will need.

• The pack, label or logo is seen in close-up: model it in CGI
• The product is not manufactured yet: CGI from the design files
• The film needs a place, people or weather: AI for the world, CGI for the product
• You need many versions to test: AI, with the product checked in each one
• The visuals will be reused season after season: CGI, because the model is kept
• The idea is a mood or a story: AI may be enough