---
title: "AI video prompt for product ads: a fill-in template"
canonical: https://withicarus.com/blog/ai-video-prompt-for-product-ads
site: Icarus
description: "A fill-in-the-blank AI video prompt for product ads: the product, one motion, one camera move, the setting and the length, with examples by product type."
---

# AI video prompt for product ads: a fill-in template

A fill-in-the-blank AI video prompt for product ads: the product, one motion, one camera move, the setting and the length, with examples by product type.

_October 7, 2026. Written with AI from the sources listed below, and checked automatically before it went live._

![Frosted glass serum bottle with a dropper on a wet stone ledge in soft morning bathroom light](https://icarus.b-cdn.net/blog/ai-video-prompt-for-product-ads/serum-bottle-stone-ledge-morning.png)

_Made with Icarus_

A good AI video prompt for a product ad covers five things: the product, one thing that moves, one camera move, the setting and the length. Write them as two or three plain sentences and start from a real photo of your product, and the clip will look like the thing you sell.

This post gives you the template, shows you how to pick a camera move from a list instead of describing one, and works through examples for skincare, shoes, jewelry, bags and clothing. It ends with what to change when the product in your clip stops looking like yours.

**Key takeaways**

- Fill five slots: product, motion, camera, setting, length.
- Ask for one action and one camera move per clip.
- Start from a photo of the real product, not from words alone.
- If the product drifts, cut the motion before you add more words.

## The five-part template

Every product ad prompt answers the same five questions. Fill in the blanks and you have a prompt.

| Slot | What you write | Example |
| --- | --- | --- |
| Product | What it is, in a few words | A glass serum bottle with a dropper |
| Motion | The one thing that moves | A drop falls from the dropper |
| Camera | One move, picked from a list | Slow push in |
| Setting | Where, when, what light | A bathroom shelf, soft morning light |
| Length | How long the clip runs | 5 seconds |

The slots come from how video models read a prompt. Google's Veo guide calls the subject [the "who" or "what" the video revolves around](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/video/video-gen-prompt-guide), and it describes actions as [the "verb" of your video](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/video/video-gen-prompt-guide). The same guide says the scene covers [the "where" and the "when"](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/video/video-gen-prompt-guide). For an ad, the product is your subject. Everything else is there to sell it.

The full template reads like this:

> [Product] in [setting and light]. [One action]. Camera: [one move]. [Length].

Keep it short. OpenAI's prompting guide for Sora 2 says [shorter prompts give the model more creative freedom](https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide). In an ad you want less freedom with the product and more with the background, so spend your words on the product and the action and let the light and props fill in.

## Pick the camera move from a list

Most people get stuck on the camera slot. You don't have to write it as film language. Pick one move and name it.

The Veo guide says camera movement [helps introduce dynamism into the shot](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/video/video-gen-prompt-guide). The Sora guide goes further: [each shot should have one clear camera move and one clear subject action](https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide). Two moves in one short clip usually means neither one lands.

In Icarus, on the video models that offer it, you open Camera and pick from a list: push in, pull out, slide left, slide right, rise up, lower down, hold still or shift focus. Leave it on Auto and the model picks. Here's what each one does for a product:

- **Push in:** moves toward the product. Good for texture, labels and the moment someone looks closer.
- **Pull out:** starts tight and reveals the scene. Good for showing where the product lives.
- **Slide left or right:** glides past the product. Good for a row of colours or a shelf.
- **Rise up or lower down:** moves over or under it. Good for tall items like bottles and boots.
- **Hold still:** the camera stays put and only the product or scene moves. The safest choice when the product must look exact.
- **Shift focus:** moves sharpness from the foreground to the product. Good for jewelry and small items.

Framing matters too. The Sora guide notes that [a wide shot from above will emphasize space and context](https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide), while a close-up pulls attention onto feeling. Product detail ads want the close-up. Lifestyle ads can go wide.

You'll find the camera list on [photo to video](https://withicarus.com/features/photo-to-video).

## Worked examples for common products

Here's the template filled in for five product types. Swap in your own product and setting.

**Skincare bottle**

> A frosted glass serum bottle on a wet stone ledge in a bright bathroom, soft morning light through a frosted window. A single drop falls from the dropper back into the bottle. Camera: push in. 5 seconds.

**Running shoes**

> A white running shoe on a damp city pavement at dawn, low warm light. A foot steps down into the shoe's frame and pushes off. Camera: hold still, low to the ground. 6 seconds.

**Necklace**

> A thin gold chain necklace on a woman's collarbone, warm window light in a quiet bedroom. She turns her head slowly to the side and the pendant catches the light. Camera: shift focus from her shoulder to the pendant. 5 seconds.

**Leather tote**

> A tan leather tote bag hanging on a café chair, late afternoon sun on a terrace. A hand reaches in and lifts out a notebook. Camera: slide right. 6 seconds.

**Linen dress**

> A woman in a sage linen dress walking along a sunlit beach path, a light breeze. The hem moves as she walks toward the camera. Camera: pull out. 8 seconds.

Every example has one action and one move. None of them asks the product to spin all the way round. A full turn shows sides your photo never had, and the model has to make those sides up.

## Start from a photo of your product

Words alone can't hold your logo, your stitching or the exact shade of your bottle. A photo can. In image-to-video, Google's docs say [Veo uses the input image as the initial frame](https://ai.google.dev/gemini-api/docs/veo), so the first second of your clip is your real product, and the prompt only has to describe what happens next.

That changes how you write. When you start from a photo, drop most of the product description and write about the motion and the camera instead. The photo already shows what the product looks like.

If you don't have a good still yet, make one first. In Icarus, product pinning keeps your real colour, logo and stitching in the image, and any image you make can become a short video in one step. A starting still might read:

> A frosted glass serum bottle with a dropper on a wet grey stone ledge, soft morning light from a frosted bathroom window, eye-level close-up

Try this prompt in Icarus: https://withicarus.com/dashboard?prompt=A%20frosted%20glass%20serum%20bottle%20with%20a%20dropper%20on%20a%20wet%20grey%20stone%20ledge%2C%20soft%20morning%20light%20from%20a%20frosted%20bathroom%20window%2C%20eye-level%20close-up

Then open that image, choose video, pick a camera move and add one line of motion. You can read more about how [your exact product](https://withicarus.com/features/exact-product) is kept in each shot.

## What to change when the product drifts

Sometimes the label blurs, the cap changes shape or the logo melts halfway through. Don't add more words. Take things away.

The Sora guide says [the model generally follows instructions more reliably in shorter clips](https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide). Shorter clips and smaller moves give the product less chance to change. Work down this list and try again after each change:

- [ ] Start from a sharp photo of the real product, front-facing, one angle
- [ ] Cut the action to one small movement
- [ ] Switch the camera to hold still or a slow push in
- [ ] Delete product description from the prompt and let the photo carry it
- [ ] Make the clip shorter
- [ ] Keep hands and props away from the label
- [ ] Move the motion to the scene instead: steam, light, fabric, water

That last fix works well. A still bottle with light moving across it often sells better than a bottle flying through the air, and the label stays readable. Check every clip against the real product before you run it as an ad.

Testing costs little if you draft small. On the AI we start you on, a 10-second clip is $2.35 at 480p, $5.08 at 720p and $12.48 at 1080p. Three drafts at $2.35 come to $7.05, so test at 480p and make the final at the size you need. Every price is on the [pricing page](https://withicarus.com/pricing), and a clip that fails costs you nothing.

To try it, make one still of your product, then turn it into a clip with one camera move from the list.

## Questions

### What should an AI video prompt for a product ad include?

Name the product, one action, one camera move, the setting and light, and the length. Keep it to two or three plain sentences. If you start from a photo, describe the motion more than the product.

### How long should an AI product video be?

Keep it short. OpenAI's Sora 2 guide says models follow instructions more reliably in shorter clips, and a short clip gives the product less time to change. Make several short clips and cut them together if you need a longer ad.

### Can I make a product video from a photo?

Yes. Most video models can start from an image, and Google's Veo docs say the input image becomes the first frame. In Icarus, any image you make can become a short video in one step.

### Which camera move works best for a product ad?

Hold still or a slow push in keeps the product most stable, so they're good choices for labels and logos. Pull out suits lifestyle scenes, and shift focus works for small items like jewelry. Pick one move per clip.

## Sources

- [Sora 2 prompting guide](https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)
- [Veo on Vertex AI video generation prompt guide](https://docs.cloud.google.com/vertex-ai/generative-ai/docs/video/video-gen-prompt-guide)
- [Veo 3.1 - Gemini API docs](https://ai.google.dev/gemini-api/docs/veo)

---

Site: Icarus (https://withicarus.com)
Canonical: https://withicarus.com/blog/ai-video-prompt-for-product-ads
Contact: tanmay@onai.studio