How to make an AI commercial from a product photo

A 15-second AI commercial from one product photo costs $1.35 to $3.60 per take at Runway list prices (checked 2026-10-26). This page gives the six-step workflow, the prompts, the cost table and a scorecard to pick the model that keeps your label intact.

How do you make an AI commercial?

You make an AI commercial in six steps: write the offer, storyboard three shots, photograph the product, animate each shot from that photo, score the results and edit them together.

1. Offer. One sentence: what it is, who it is for, one proof point.
2. Storyboard. Three 5-second shots make a 15-second ad. Hook, product in use, product close-up with the call to action.
3. Product photo. Shoot the product from the front with the label flat and sharp, and check the photo before you spend credits. Use the largest, cleanest file you have.
4. Animate from the photo. Use image-to-video on a model that accepts a start frame. Confirm that in the model's own documentation before you pay for it.
5. Score every take. Use the scorecard below.
6. Edit. Cut the kept shots, add captions, music and the call to action in your normal editor.

What do other AI commercial guides miss?

The three competing guides reviewed explain the steps but give no measured label retention, no failed shots and no model-specific, reproducible generation-cost calculation with retry scenarios.

Wireflow says image-to-video preserves labels, gives no retention measure and prices a 30-second ad as "single-digit to low double-digit dollars" without naming a model. Renderforest gives a nine-step workflow and no AI tool prices. YOPRST quotes "$1,000 to $5,000" per AI production and does not break it down. Checked 2026-10-26.

How much does an AI product ad cost?

At Runway API list prices, one take of a 15-second ad built from three 5-second shots costs between $1.35 and $3.60 for silent models, before retries.

Runway sells credits at $0.01 each (checked 2026-10-26); see Runway pricing for how its credits work. Cost per take is shots × seconds × credits per second × $0.01. Veo 3.1 on Runway generates 4, 6 or 8 seconds per clip and accepts 1080p only at 8 seconds, so the Veo rows use three 6-second clips at 720p, each trimmed to 5 seconds in the edit.

Arithmetic for Gen-4.5: 3 shots × 5 s × 12 credits = 180 credits = $1.80. The last two columns price the case where the 3rd or 6th take is the one you keep.

Add $0.02 for each start-frame still if you generate it with Gen-4 Image Turbo. Your own product photo costs nothing. Budget every retry as a full charge.

Runway also lists Kling 4 at 15.7 credits per second at 720p and 23.5 at 1080p, with a 20% discount through 30 November 2026 UTC. Prices change often, so check the page before you set a budget.

How do you keep the label and logo from drifting?

Start each shot from a real photo of the product and keep camera motion slow, so label text has the least to follow.

• Use image-to-video, not text-to-video. Text prompts have no photo to copy.
• Write the product as a fixed object: "the bottle stays upright and the label faces the camera."
• Put the logo in the editor as an overlay for the end card. Do not ask the model to render it.
• Keep the label large in frame in the close-up shot, and check small print at full size after every take.
• Reject any take where the text changes between the first and last frame.

What prompts work for each shot?

Each prompt should name the start frame, one movement, one camera move and one thing that must not change.

Replace the scene details with your own product.

How do you score a take?

Score each take on label drift, shape and edit time so that you can compare models on your own product with numbers.

Keep the failed takes. They show which shot type breaks on which model.

Can you run AI video ads on Meta and Google?

Yes. Disclosure rules differ by platform and ad type, and ordinary product ads face the lightest ones.

Meta's June 1, 2026 update says it will begin automatically detecting ads created or edited with third-party AI tools through industry-standard signals, and show an "AI info" label in the "About this ad" menu. Meta also requires advertisers to disclose digitally created or altered image, video or audio in social issue, election and political ads.

Google's July 2024 policy requires the "Altered or synthetic content" checkbox for election ads with synthetic content that depicts real or realistic-looking people or events. In mobile feeds, mobile Shorts and in-stream formats Google generates the in-ad disclosure from the checkbox. In other formats you add your own clear and conspicuous disclosure.

Do not use a real person's face or voice, or a copyrighted character, without permission. Check the claims in your script against what the product actually does.

Should you use an AI video ad generator instead?

An AI video ad generator bundles the steps above into one tool. It earns its place if it accepts your product photo, exposes shot controls and lets you regenerate individual shots.

Use one when you need many variants fast. Use the manual workflow when the packaging must be exact. Either way, apply the same scorecard.

For model prices beyond this table, see Kling AI pricing and Sora alternatives. The Learn hub lists the rest, and HoneyVid is where you pick a model and create.