# How to make AI product ads with Veo

> Make AI product ads with Veo: give it photos of your product, write one move and one sound per clip, then edit the best clips. Includes Google's 2026 changes.
- **Author**: Marxx
- **Published**: 2026-11-17
- **Category**: AI
- **URL**: https://marxx.ai/library-guides/how-to-make-ai-product-ads-with-veo

---

To make an AI product ad with Veo, give it reference photos of your real product, write one action and one sound per clip, then edit the best clips into an ad. You can use Veo in the Gemini app or Google Flow, or through the Gemini API if you code. Google's own pages change fast, so read the notes below first.

## What Google says today (checked 2026-10-20)

- Google DeepMind calls Veo 3.1 its leading video model. It generates sound natively: effects, ambient noise and dialogue. It accepts reference images of a scene, character or object, and supports first and last frame control. It outputs 1080p and 4K, and it marks videos with SynthID.
- For developers, Google says the Veo 3.1 preview models in the Gemini API shut down on 22 October 2026. It points developers to Gemini Omni Flash, and calls that its default video model. It keeps Veo 3.1 for features such as scene extension and last-frame control.
- Google bills Gemini Omni Flash by output tokens. It says that works out to about $0.10 per second of 720p video under standard pricing.

If you use a no-code tool, the steps below stay the same. The tool name may change.

## Step 1: choose where to work

Try Veo in Gemini or Google Flow if you do not code. Use the Gemini API if you want to build a repeatable workflow.

## Step 2: gather references

Upload two or three sharp photos of the real product from different angles. Google says Veo takes reference images of an object to guide the video, which helps keep your product's look steady from clip to clip.

## Step 3: write one move and one sound

Describe the subject, one action, one camera move and one sound. Veo makes audio, so say what you want to hear.

> A steel water bottle on a kitchen counter. The hand twists the cap open, then tilts the bottle to pour. Slow push-in. Sound: the soft click of the cap, then water pouring.

## Step 4: control the start and end

Use first and last frame control for a planned change, such as a closed box and then an open box. It keeps the clip from drifting.

## Step 5: check every frame

Pause on the first, middle and last frame. Look for a changed label, a warped shape and odd hands. Delete clips that fail. See [how to avoid AI-looking ads](/guides/how-to-avoid-ai-looking-ads).

## Step 6: edit and add real text

Cut three or four clips together. Add the price, offer and logo as real text in your editor. Veo's outpainting can expand a clip to fit a different screen shape, but check the edges.

## Step 7: label and review

In India, ASCI's labelling guidelines take effect three months after their 29 September 2026 release. They require a label when you show a product that does not currently exist. Read [how to label AI-generated ads on Meta](/guides/how-to-label-ai-generated-ads-on-meta).

## A worked example

Say you want five 8-second shots for a INR 1,299 bottle and you expect three tries per shot. That is 5 x 8 x 3 = 120 seconds of video. At about $0.10 a second, the API cost is $12. If six clips are usable, you pick the best five. Compare that with the cost of a shoot, then test the result against a real video. Prices change, so check Google's pricing page on the day you buy.

## Mistakes to avoid

- Reference photos that are dark or low resolution
- Asking for more than one action per clip
- Leaving the audio line out
- Forgetting that preview models can shut down
- Not checking the label in every frame

For a side-by-side with another model, see [Kling vs Veo](/compare/kling-vs-veo). We make Marxx, whose [production canvas](/features/production-canvas) generates images and video. It builds the ad in Meta, always created paused.

## FAQ

### Does Veo make sound?

Yes. Google DeepMind says Veo generates audio natively, including sound effects, ambient noise and dialogue.

### Is Veo available through the Gemini API?

Yes, but Google says the Veo 3.1 preview models there shut down on 22 October 2026 and points to Gemini Omni Flash. Check the current Gemini API video page before you build.

### Can I upload my product photo?

Yes. Google DeepMind says Veo accepts reference images of a scene, character or object to guide a video.

---
- [More AI articles](https://marxx.ai/library-guides/category/ai)
- [All articles](https://marxx.ai/library-guides)