AI video demos can make professional-looking production seem almost frictionless: type a prompt, wait a few seconds, and a finished shot appears.
The production reality is much messier.
Business Insider reported that the creators of the AI-generated sci-fi series The Arc used roughly $38,000 in AI credits and averaged eight to ten generations for each usable shot. The five-episode project was completed in a little over three weeks using a mix of generative tools and human creative direction.
Those numbers are useful because they expose the hidden work behind polished AI video.
AI removes some production costs and creates different ones
A generative project may avoid location rentals, physical sets, travel, wardrobe, and certain crew costs.
That does not make production free.
The budget shifts toward generation credits, repeated attempts, prompt development, continuity management, asset organization, compositing, editing, sound design, and review.
The labor changes shape rather than disappearing.
This distinction matters for businesses budgeting AI campaigns. A company that assumes “no camera crew” automatically means “cheap” can underestimate how much iteration professional work requires.
Eight to ten generations per shot is not a failure
Generative video is probabilistic.
A director may know exactly what the shot needs to communicate and still receive outputs with the wrong movement, character appearance, framing, lighting, wardrobe, emotion, or continuity.
The first generation is often exploration.
The next several are refinement.
The usable shot may come late in that process.
That makes generation count a production variable, much like takes on a traditional set.
PHQ expert take: budget AI around iteration, not around prompts
A professional AI-video estimate should not be based on the number of final shots alone.
It should account for how difficult those shots are to control.
A simple abstract transition may require only a few attempts.
A recurring character performing specific actions across several scenes may require far more.
That means the creative brief matters enormously.
The more specific the brand, continuity, product, or character requirements, the more room should exist for iteration.
Human direction remains the scarce skill
AI can create options quickly.
Someone still has to decide which option belongs in the final piece.
That person needs to understand story, composition, brand consistency, pacing, audience, and quality.
As generative tools improve, those judgment skills may become more valuable rather than less.
If everyone can produce an image, taste becomes the differentiator.
If everyone can generate video, direction becomes the differentiator.
Hybrid production will often make more sense
AI does not need to replace conventional production to be valuable.
Many commercial workflows can combine both.
A brand might record a real founder or spokesperson, then use AI for environments, cutaways, visual metaphors, concept scenes, or transitions that would be expensive to film.
A podcast team could use real studio footage for the trusted human core and generative visuals for supporting explanations.
That can create a better balance between authenticity and production flexibility.
Where AI video is especially useful
Generative video can be compelling for ideas that are difficult or expensive to capture traditionally:
- concept trailers,
- stylized commercials,
- surreal environments,
- visual explainers,
- previsualization,
- historical recreations,
- branded entertainment,
- and campaign extensions.
It is less compelling when a simple camera shoot would be faster, more reliable, and more authentic.
The correct question is not “Can AI make this?”
It is “Which production method gives us the best result for the budget and timeline?”
Production management still matters
Large AI projects create their own asset-management problem.
Teams can end up with hundreds or thousands of generations.
Without good naming, versioning, reference files, approvals, and shot tracking, the project becomes difficult to manage.
That is familiar production work in a new form.
Professional teams need systems around generative media just as they need systems around camera footage.
The PHQ perspective
The $38,000 credit figure is not evidence that AI video is too expensive.
It is evidence that professional AI production is still production.
Good work requires iteration, direction, organization, and quality control.
Creators and brands should use AI where it expands what can be made or materially improves efficiency.
But they should budget it honestly.
The future is not one-click filmmaking.
It is a larger production toolbox—and teams that know when to use each tool.
Sources
Primary reporting and supporting material used for this article.
