How Much Does an AI Video Ad Cost? Quote Per Shot, Not Per Second

If a supplier quotes your AI spot per second, your budget may run out halfway through production. Anyone asking how much an AI video ad costs usually gets one simple figure: a set price for every second of the spot. It is easy to compare. But it does not describe the work, because it hides the variable that matters most: how many attempts it takes to get the right shot.

Our position is straightforward. An AI ad should be priced per shot, with a stated estimate of attempts for each one. The second is the unit of the finished product, not the unit of the work.

The per-second price measures the tool, not the work

Model pricing really is per second, and it moves fast. On 5 November 2026 Google released Gemini Omni 1.2 Flash, and its list price is $0.07 per second of 720p video, about $1.05 for a 15-second clip, a 30% cut from the previous version.

That figure is useful, but it says one thing: what it costs to generate one second of video with that model at that resolution. It does not say how many seconds you will generate to end up with five good ones. It does not say who reviews the results, who corrects them, who edits, who delivers. A quote that takes that number and multiplies it by the length of the spot measures a tool's price list, not the cost of an ad.

Every shot has its own curve of attempts

A 30-second spot is not a single block. It is ten, twelve, fifteen shots, and they are not equal. A landscape with the right light may land in a few attempts. A hand picking up the product, a legible logo in motion, the same face across six scenes all take many more tries.

An example, with hypothetical numbers for illustration only. Take a 30-second spot made of 10 shots of 3 seconds each, at a list price of $0.07 per second:

• 6 simple shots, 5 attempts each: 90 seconds generated.
• 3 standard shots, 15 attempts each: 135 seconds generated.
• 1 critical shot (product in hand, legible logo), 40 attempts: 120 seconds generated.

That is 345 seconds generated for 30 seconds delivered, more than 11 seconds of generation for every second of spot. One critical shot weighs almost as much as the three standard ones. A per-second quote would treat them all alike.

In this scheme, the raw generation cost stays a modest line. The real weight sits around the attempts: direction time to write and correct each prompt, review time to discard results, post-production to make the chosen shots consistent. For more on the per-shot logic, see our case on AI video generation cost per shot.

One practical detail: models let you set the length of each clip. With Omni 1.2 Flash, you can set anything from three to fifteen seconds in one-second steps. Generating exactly the length a scene needs avoids paying for seconds the edit will cut.

A per-shot quote has five lines

The quote we recommend asking for (from us or anyone else) is built like this:

1. Shot classification. Every storyboard scene is simple, standard or critical, with the reason.
2. Expected attempts. An estimate per class, written in the document, with a ceiling beyond which both sides talk.
3. Direction and review. The time of the person directing the prompts and selecting results, and the review rounds included.
4. Post-production. Editing, colour grading, sound, upscaling, consistency across shots.
5. Delivery. Formats, lengths and versions required (16:9, 9:16, 15-second cut-downs).

Lines 1 and 2 separate an honest quote from an optimistic one. Without them, the risk of underestimating attempts sits entirely with the client and arrives as extras at the end of the project.

The tool changes every month, the quote structure does not

A 30% cut in a single release shows how much list prices move. A quote anchored to a per-second rate goes stale with the next version. A per-shot quote separates two things: the structure of the work, which stays stable, and the model price, which is one line you can update.

The same goes for choosing a model. Different models have different strengths in motion, references and consistency, so the same scene may need a different number of attempts. When comparing suppliers, the attempt estimate matters as much as the per-second price. For help choosing, see our guide to the best AI video models of 2026.

How we build the quote for a spot

We start from the brief and the storyboard and split the spot into shots before we talk about figures. On the scenes we expect to be hardest, we run generation tests before closing the quote, so the attempt estimate rests on trials, not on impressions. The document we deliver lists shots by class, the review rounds included, and what happens if a scene needs more tries than planned. The full path from brief to delivery is described on our AI Video Production page.

What to ask a supplier before you sign

Four questions are enough to tell whether a quote is solid:

• How are shots classified, and which count as critical?
• How many attempts are estimated per class, and what happens beyond the ceiling?
• How many review rounds are included, and on what?
• Which formats and versions are in the price?

To judge whether the spend pays back, we wrote a guide on how to measure the ROI of AI content, and for the whole production path see how an AI commercial is made. If you have a storyboard or a brief and want a per-shot quote, get in touch via the AI Video Production page.