An AI Ad for a Restaurant, a Café or a Dish

Quick answer: Two kinds of restaurant ad: the room, or the food. Lead with a photo of the dining room or the shopfront and pick A day in the life, and the ad is people enjoying your place — your actual room, with the customers in your photo removed and a generated cast in their seats. Lead with a photo of the dish and pick Mouth-watering close-ups, and the ad is steam, pour and cut. Add the other kind as a second photo. Fifteen seconds is enough for one dish; thirty for the room and a dish. The end card carries the name from your sign.

Still from “Crimson Vow”, an AI short drama generated end-to-end by SceneMixer (4 characters, 1 episode) — Sample: “Crimson Vow” — Realistic HD, 1 episode, produced end-to-end by SceneMixer from the script · watch the episodes

Which photos work and the four pickers cover the general rules; this page is the restaurant case.

Which photo first

The first photo is what the ad is about. The room or the shopfront first when the draw is the place: the reading files it as a store / property, keeps your actual interior or sign, removes the people, and sets the ad inside it. The dish first when the draw is the food: it is cleaned up into a product shot and the ad is built around it. Add the other as the second photo and both appear.

The episode asset library showing generated character sheets, location images and prop images — A restaurant photo becomes the place the ad happens in; the dish becomes the featured product.

Taking the photos

Room: stand at the door and shoot the whole space in daylight; include the sign if the name matters. Dish: fill the frame, top-down or three-quarter, on the table you actually serve it on. Customers and staff in the room photo are fine — the composer says People will be removed. A menu photo is not a good first image; it reads as text.

Which vibe

Mouth-watering close-ups for the food: heat, pour, the first cut, the bite. A day in the life for the place: a couple at a window table, a family, the evening rush. Warm for a family restaurant; Minimal for a fine-dining plate; Funny for a street-food stall with a personality.

Length and format

Fifteen seconds: one dish, one idea. Thirty: the room, then the food. Sixty: a small story — arrival, the meal, the goodbye. Portrait for feeds and stories; landscape for a website or a screen in the window.

The one line

Type what it is and who it is for: our family-run Sichuan restaurant, for people who like it spicy. It tells the reading this is a place, not a product, and gives the script its angle. Skip health claims and superlatives; they are removed anyway.

The end card

The name on your sign is read and printed as the brand; the product name is what you confirmed in the dialog. Add a tagline and a call to action — Book a table, Order on the app. The name is delivered here, not on the shots. The end card.

What to expect

The dishes look like your dishes; the room looks like your room; the diners are a generated cast that fits the place. Lines are spoken in the dialogue language shown in the composer. If a shot is wrong, open it in the editor, change the text, regenerate that shot, and rebuild the final cut. The progress page.

You are selling	First photo	Vibe	Length
A signature dish	The dish, filling the frame	Mouth-watering close-ups	15s
The place	The room or the shopfront	A day in the life / Warm	30s
Both	The room, then the dish	A day in the life	30–60s
A café’s drink	The drink, poured	Mouth-watering close-ups / Minimal	15s

FAQ

My restaurant photo has customers in it.

Fine. People are removed and the room is kept; the ad's own cast sits in it.

Will the ad show my actual dining room?

Yes, when the room is your first photo. It is drawn from your photo, without the people.

Will the dish look like my dish?

Yes. A dish photo is cleaned into a product shot and the ad is built around it.

Does the restaurant's name appear in the shots?

Not reliably; generated video does not write legible text. The name is delivered on the end card.

Can I say the food is the best in town?

Superlatives and health claims are removed from ad scripts. Say what it is and who it is for.

One dish, one photo, one ad

Add the photo, pick Mouth-watering close-ups, and make the ad.

Make an ad

Updated September 26, 2026