Frame from the Haval Dargo commercial
AI production · Generative video

AI production: commercials made with generative models

Generative models now deliver shots that used to need a location, a crane and a shooting day. We make the whole commercial inside them — script, storyboard, colour and sound — with the same directing and the same quality control as a shoot.

Discuss a project ↗ Portfolio ↗

How we approach AI video

A model is a tool, not a replacement for direction. The film starts where a shoot starts: with the brand’s objective, an idea and a script. The model stands in for the location, and it shapes the result about as much as the choice of camera and glass.

The hard part of generative video is holding things still between frames: a face, a pack colour, the grain of a material, the direction of the light. That is why the visual language and the key frames are agreed before generation, not discovered during it.

Assembly runs the usual way: editing, colour, graphics and sound are handled by the team that runs post-production on our shoots. The viewer should not be able to tell how the film was made.

What it solves

  • Shots that are expensive or impossible to film: unreachable locations, large-scale scenes, stunts no one can perform.
  • Deadlines that leave no room for prep and a shooting day.
  • High visual ambition against a limited production budget.
  • Testing several creative directions before a brand commits to a full shoot.

What’s included

Idea and script

01

A concept built around the brand’s objective, and a script written around what generative models do well — and what they still do badly.

Storyboard and style

02

References, visual language and key frames, all signed off before a single shot is generated. Reworking finished footage costs more than reworking a board.

Generation

03

Shot-by-shot work in generative models, holding characters, objects and light consistent from one frame to the next.

Assembly

04

Editing, colour, graphics, sound and voice — the film is finished to broadcast standard by the same people who finish our shoots.

Versions and delivery

05

Durations and formats for broadcast and platforms, revisions, final masters.

Who it’s for

  • Brands that need imagery a conventional shoot cannot reach.
  • Agencies working against a campaign deadline.
  • Brands testing ideas before committing to full production.

Why Firmachi

  • Production discipline: script, storyboard and sign-off happen before generation, so the result is predictable.
  • Eight years of commercial shoots — we know the difference between a pretty picture and a frame that works for the brand.
  • Post-production in-house, so AI frames and filmed material end up in one visual style.
  • If part of it is better filmed, we will say so: a commercial and AI scenes can live in the same project.

FAQ

How is an AI film different from a shoot?
No location, no lighting package, no shooting day, so the schedule and the production budget come down and the scenes can be anything. In exchange you get less control over fine detail: if exact product rendition matters, some shots are better filmed.
Can AI and filmed footage be combined in one film?
Yes, and it usually turns out better that way: we film the product and the cast, and generate the environment, the wide shots and the unreachable locations. Post-production brings it into one look.
What does an AI film cost, and how long does it take?
It depends on duration, the number of scenes and how much detail is required. We fix the quote and the schedule after the brief, exactly as we do on shoots.

Let's discuss your project

Tell us about the task and timing. We'll reply within one working day and suggest the next step.

Discuss a project ↗
Other services