Guide · Cost
How much does an AI commercial cost? An honest look.
Not a price list, but what price lists leave out: the seven factors that actually drive the cost of an AI commercial, with real numbers from our own productions and the industry.
Last updated: July 2026
The short answer: an AI commercial costs significantly less than a comparable live-action shoot, because set, cast, location, gear and travel disappear. It is still not free, for one simple reason: the price is driven by the working hours of people who know what they are doing, not by tool subscriptions. Anyone quoting a fixed price before knowing your project is guessing.
Why “AI is basically free” is a myth
Generation costs really are small. When prediction market Kalshi had its viral 2025 NBA Finals spot made with Google Veo 3, pure prompting costs stayed under 2,000 dollars. The more interesting part: it took roughly 300 to 400 generations to get 15 usable clips. A rejection rate above 95 percent is the norm, not the exception.
From our own production: the Manner spec spot was made in about ten hours. That sounds fast, and it is. But those were ten hours with no client, no feedback loops, no brand guidelines and no rights clearance: the prototype of a spot, not a client project. As soon as a real brand is involved, the time-consuming parts arrive.
The seven cost factors
Factor 01 · Concept and script
The idea is 90 percent of the work. A spot with a real story needs a script, storyboard and references before the first frame is generated. This is where advertising happens, or random AI imagery.
Factor 02 · Length and number of scenes
Scenes cost money, not seconds. Ten different scenes in 20 seconds are more work than one continuous shot over 60, because every scene means its own generations, its own consistency checks and its own rejects.
Factor 03 · Consistency requirements
The biggest price driver. A recognisable character across shots, a product that must look exactly like the one on the shelf, a brand world with fixed colors: consistency is built as a system of reference frames, start and end frames and many iterations, not in a prompt.
Factor 04 · Formats and versions
16:9 for YouTube, 9:16 for Reels, 6-second cutdowns for ads: every version is editing and often generation work. Defining formats early avoids paying twice.
Factor 05 · Sound
Voice-over, music licensing, sound design and the mix turn images into a commercial. Good sound is chronically underestimated and is its own step with its own licensing costs.
Factor 06 · Feedback loops
Every revision round means new generations, rejects included. Two fixed feedback rounds with consolidated notes are cheaper than ten individual wishes over weeks.
Factor 07 · Rights and usage
Commercial usage rights of the models, music licensing depending on where the spot runs (social, TV, cinema) and the license period need clean answers. Cheap when planned from the start, expensive when not.
What this means for your budget
Instead of a price list that would be wrong for your project anyway, the reverse works better: define a rough budget range and the goal of the video. From that, it is possible to say honestly what is achievable within that range, which level of consistency is realistic, and where AI saves the most compared to live action. You usually get an assessment within 24 hours: get an AI commercial made.
Not sure whether to outsource at all? Read AI video: DIY or hire a pro? first.
FAQ
Frequently asked questions
- Is an AI commercial cheaper than a live-action shoot?
- In most cases significantly, because set, cast, location, gear and travel disappear entirely. The savings are biggest for scenes that would be expensive or impossible in reality: exotic locations, time travel, stunts, historical settings, or products that don't physically exist yet. For context from practice: our Manner spec was made in about ten hours, whereas a comparable live shoot would have needed shoot days, a team and a location. Where real faces and emotion matter, though, a live shoot is often still the better choice, and that's exactly where the hybrid approach pays off. So an AI commercial is almost always cheaper, but not automatically the right choice for every subject.
- Why is there no price list?
- Because the price depends almost entirely on concept, number of scenes, consistency demands and formats. Two 30-second spots can differ in effort by a factor of five: one continuous shot with a single product is something completely different from ten scenes with a recurring character who has to look the same everywhere. A fixed number on a price list would therefore be wrong for your specific project almost every time. The reverse works better: you name the goal and a rough budget range, and I tell you honestly what's realistic within it and where AI saves the most. That leads to a reliable answer faster than any flat rate, usually within 24 hours.
- What makes an AI video more expensive?
- Mainly three things. First, consistency: recurring characters or products that have to look exactly the same across many shots are the biggest price driver, because they need a system of reference images and many iterations. Second, the number of scenes: ten different scenes cost more than one long shot, because each scene means its own generations and its own rejects. Third, the number of format versions for YouTube, Reels and paid ads. On top of that come music licences depending on where it runs, and sound overall, which is chronically underestimated. It gets cheaper the other way round through clear concepts, a few strong scenes and formats defined early.
- How fast do I get a cost assessment?
- With a goal, a rough budget range and a timeline in your inquiry, usually within 24 hours. The more concrete your details, the more reliable the assessment: one sentence on the idea, the approximate length, where the spot will run and a rough range are enough for a first honest answer. From that it's possible to derive what effort is realistic, whether fully generative or hybrid makes more sense, and where AI saves the most compared to a live shoot. If a project doesn't fit the range you name, I'll say so openly instead of massaging a number. That way you don't waste time on quotes that were never going to match.