Services

AI tools

The models behind the films — and what each one is for.

Overview

AI video changes every few months. This page lists the tools used in my commercial work today, what each one is good at, and how they fit together with traditional motion design and 3D. Last updated October 2026.

No single model wins every shot. The film is planned shot by shot, each shot goes to the model that suits it, and everything is finished in the same pipeline as filmed work — edit, compositing, sound and grade.

How a film is made

01

Script and shot list

The film is written and broken into shots before anything is generated.

02

Keyframes — Nano Banana Pro

A designed still for each shot, with the product and characters locked.

03

Shots — Seedance, Veo, Kling in Higgsfield

Each keyframe turned into motion with the model that suits the shot.

04

Product accuracy — Cinema 4D and compositing

Labels, logos and geometry fixed where a model cannot hold them.

05

Finishing — After Effects, edit, sound, grade

Titles, end cards, music, sound design and colour.

The tools

01

Higgsfield

One studio for Seedance, Veo, Kling and other models, with Cinema Studio for camera and lens control and Soul ID for consistent characters. Used to direct whole sequences in one place.

02

Seedance 2.0

ByteDance’s video model. Strong at keeping products, logos and on-pack text intact and at following a detailed shot brief — the main model for product and character shots.

03

Google Veo 3.1

Realistic motion, light and physics, with native audio. Used for shots where live-action realism matters most.

04

Kling 3.0

Fluid human motion and longer, stylised shots up to 4K. Used for movement-heavy and stylised scenes.

05

Nano Banana Pro

Google’s image model. Used for keyframes, storyboards and still edits that keep the same product or character from frame to frame.

06

After Effects

Motion graphics, compositing, titles, end cards and finishing.

07

Cinema 4D & Redshift

Modelling, lighting and rendering for 3D product shots and anything that has to be geometrically exact.

The wider landscape

Other models matter too. Runway Gen-4.5 is strong on camera control and video-to-video editing, Luma Ray3 on atmospheric image-to-video, MiniMax Hailuo on fast short-form, and Wan 2.6 is an open model that can be self-hosted. For images, Midjourney V8, FLUX.2 and GPT Image 2 sit alongside Nano Banana Pro. OpenAI closed the Sora app in April 2026 and its API in September 2026.

Questions

Which AI video model is best for commercials?

There isn’t one. In 2026 Veo 3.1 is usually strongest for realism, Seedance 2.0 for keeping products, logos and text intact, and Kling 3.0 for complex human motion and longer, stylised shots. A commercial normally uses more than one.

What is Higgsfield?

A platform that brings several video and image models — including Seedance, Veo and Kling — into one studio, with Cinema Studio for camera and lens control and Soul ID for consistent characters. It is widely used for social media and advertising content.

Can you produce inside our Higgsfield workspace?

Yes. For agencies and brands that already use Higgsfield, I can work in your workspace or deliver finished films — see white-label production.

Do you use AI in motion design and 3D too?

For references, backgrounds, textures and keyframes, yes. Logos, type and product geometry are built in After Effects and Cinema 4D, so they stay exact.

Can AI-generated video be used commercially?

Each tool has its own commercial terms, and they differ by plan. You get a list of the tools used in your film, so your legal team can check them.

Services

Have a brief?

Get a quote
Exhibition & spatial design ↗