Higgos Delivery Ad

A commercial for a made-up food delivery brand, built by Higgsfield as a tutorial on directing AI video. This is the finished 30-second edit from the opening of their tutorial, with music and sound, and with Higgsfield's Cinema Studio mark in the corner.

How it was made, in three stages:

1. Build reusable assets first. Character, location and prop sheets are made in Soul Cinema and corrected in Seedream 5.0 Pro. Each is approved before any scene is written, so faces, seats and the phone stay the same between shots.
2. Have Claude write the video prompts. Each prompt opens with the cast, the phone, the locations and the look, then lists the shots second by second.
3. Generate with Seedance 2.5 and fix with blocking maps.

Where the prompts are: the exact prompt for every sheet and scene is on Higgsfield's write-up of this ad, each with a copy button, along with a download of all the assets and the Claude skill that writes the timed prompts. They are Higgsfield's own text, so what follows describes what each one pins down instead of repeating it. The full tutorial is on YouTube.

Characters

Mika. A three-panel sheet on plain grey: a front portrait, then front and back outfit views cropped at the neck so the face appears once. The prompt fixes her glasses, hair, jacket, jeans and shoes, asks for neutral colour and soft even light, and rules out freckles, retouching and any text. The first result was corrected in Seedream 5.0 Pro.

Jenny. The same three-panel layout. The prompt insists on two layers of headwear, a beanie under a raised hood, because both must stay visible through the whole ad.

Alex. The third friend, given a red checked overshirt and a loose silhouette so he stays easy to recognise in moving shots.

The mariachi band. Four musicians with distinct faces, the right instruments and full-body spacing, so each can be told apart in the Mexico scene.

The cat. A face view and a body view of the grey-and-white cat that appears in the flat and in the closing shot.

Locations

The attic flat. The prompt lays out where each friend sits (armchair left, sofa centre, beanbag right) and what is behind each seat, so the room stays consistent as the camera moves.

Italy. A canal terrace rebuilt in Seedream 5.0 Pro as a table for three, with a chair on the left, right and far side, and food and glasses for each person.

Mexico. A terrace with three seats, a market on one side and a valley on the other, leaving room for the band to move.

Japan. A low sushi table for three in a tatami room, looking out on blossom and a pond.

Props

The red phone. Front, back and angled views of the one phone that carries the story from scene to scene. The scene prompts add that its screen always faces the character, never the camera.

The Higgos packaging. A matching red delivery bag and two-cup carrier for the closing product shot, which gives the made-up brand a readable identity.

Blocking

The overhead map. A simple top-down drawing of who sits where at the Japan table. It was added as a reference after characters kept swapping seats in the overhead shot.

The three scene prompts

Scene 1, the flat to Venice. Sets up where each friend sits with one wide shot, plays three hunger gags (a growling stomach, a flower that turns into a chicken leg, a mock bite at the cat), then carries a single phone tap into Venice on a moving match cut. Every shot has a time window, a lens, a camera move and the exact moment each line starts and ends.

Scene 2, Italy to Japan. Follows the phone across three locations, linking food, the band's performance and camera movement, and ends on a doorbell.

Scene 3, the pack shot. Opens on readable packaging, then pulls back to the cat eating while the friends stay soft in the background. It runs four seconds.

What the scene prompts have in common: a style header that fixes the lens range, colour, lighting and pace before the shot list; physically plausible camera moves; no background music except the band; and one rule for the phone, that only its red back ever faces the camera.

The prompt shown here was written afterwards from the finished clip, so it describes what is on screen.