The cost of producing a clip: what a $45 job actually costs you
You know what to charge and not what you will spend — the spending is spread out and profit gets counted as "what arrived in the account". People work at a loss that way for months. Below is the second half of the equation: prices by operation and a real job broken down.
You know what to charge and not what you will spend — the spending is spread out and profit gets counted as "what arrived in the account". People work at a loss that way for months. Below is the second half of the equation: prices by operation and a real job broken down.
You know the price list and not the profit
You have looked at the market ranges and know what to charge. A short clip $18–45, an image per piece $12–35, a set of listings $245–730.
That is half the equation.
The other half is what you will spend on that job. And almost nobody counts it, because the spending is spread out: the subscription was debited a month ago, the generations run in the background, and profit gets counted as "what arrived in the account".
People work at a loss that way and do not notice for months.
Let us take the second half apart in figures.
What generation actually costs
The order of magnitude for the main operations:
| Operation | Price per generation |
|---|---|
| Image | $0.12 – 0.25 |
| Photo editor | $0.12 – 0.50 |
| Background removal | $0.12 |
| Upscale | $0.12 – 0.40 |
| Photo animation | from $0.17 |
| Video generation | $0.22 – 11.95 |
| Lip Sync | $0.12 – 1.50 |
| Music and sound | $0.12 – 1.80 |
| Transcription and recognition | $0.15 |
| 3D model | $0.36 – 0.72 |
The first thing visible: static costs pennies. Twenty generated images is a few dollars.
The second: video is the only category with a large spread. From twenty cents to almost twelve dollars per generation, depending on model, duration and resolution. That is where quoting mistakes get made.
Costing a real job
Not by one generation but by how many actually get used.
A short clip for $45
An honest cost includes the misses rather than only the successful frame.
- Script and text — pennies
- Frames: 3–4 are needed, with attempts it comes to 10–12 generations
- Animation: 3–4 frames, with redos 6–8
- Sound or narration
- Upscaling the final
Total with careful work — a few dollars. With careless work, twenty attempts per frame and an expensive video model, it can run into tens.
The difference between those two scenarios is your profit.
A set of listings for $245–730
Twenty to thirty images with corrections and upscaling — the spend still stays in the single-to-tens of dollars against a fee in the hundreds.
The conclusion on both: the cost is a few percent of the fee. The problem is not that generations are expensive. The problem is that they are not counted and not built in.
Three things that break the economics
1. Generating on the off chance
Twenty attempts instead of four, because the task was not stated.
A precise description of what is needed is not pedantry but direct saving. Plus previews and intermediate checks: do not run the final frame at maximum resolution until you know the composition is right.
2. An expensive model where a cheap one would do
The spread in video is from $0.22 to $11.95. That is a fiftyfold difference.
The rule: drafts and testing an idea on a cheap model, the final on whichever the quality requires. Running drafts on a top model is the fastest way to eat your margin.
3. Revisions with no limit
For generative video a revision often means regeneration, that is, a new cost.
"We revise until you are happy" in a verbal agreement is an open account at your expense. Two rounds of revisions in the quote is normal; beyond that is separate work.
How to build the spend into a quote
There is one rule: the cost of generations goes into the client's price the way paper goes into a printer's.
No client asks a printer what the paper cost. He pays for the print run.
The practical method. Estimate the spend for the task, multiply by two — for misses and revisions — and build it into the price. If it comes to more than five or seven percent of the fee, either the job is too cheap or the process is too wasteful.
What not to do: put the spend as a separate line on the invoice. The client will start discussing the cost of generations rather than the result, and you will be in a conversation with nothing to win.
Why per-use payment matters more than discounts
Separately, because it changes the very possibility of working.
A subscription is payment for access. It is debited the same whether you did two jobs that month or twenty. The more jobs there are in a month, the smaller the share of the subscription in the cost of each one, so it is worth counting it together with the jobs rather than separately.
A wallet is payment for work. Did not work for a week, spent nothing. A month of idleness costs nothing.
Three properties that decide it:
The price is visible before you press. You see the cost of an operation before launching rather than after. You can compare two models and choose.
Money is returned to the balance for anything not delivered. A step did not complete and its cost returned to the balance automatically.
The entry threshold is small. You can start with a sum you do not mind and then pay out of what you earn.
That is a direct answer to the objection I hear from my audience verbatim: "the high cost of AI, subscriptions are very expensive, there is no room to experiment". The room appears exactly when an experiment stops costing as much as a monthly subscription.
What to measure so you are not guessing
Three figures, and everything becomes clear within a month.
Spend per job. Not per month but on a specific piece of work. The right question is "what did this job cost me" rather than "what does the subscription cost".
The spend as a share of the fee. The job brought $40 and cost $3 — seven percent, fine. It cost $20 — the issue is not the prices but the process.
The number of attempts per successful result. The most useful figure. It shows where you are losing: in stating the task or in choosing the model.
What the platform offers for costing a video
Gen AI — 338 models in one place instead of a dozen subscriptions. The price is shown beside each. The models differ in quality and speed: different ones are taken for a draft and for the final.
Prompt Master — the "Improve prompt" button rewrites the phrasing to a method. That is direct saving: a precise request reduces the number of attempts.
Cerebrum — a chat where you describe the task in plain words and the right models are picked automatically. The estimate is visible before launch: what a move costs and what will be left on the balance. Paid from your balance as used.
Registration is free and opens three days of full Basic access. Express courses on image generation and video models — "Midjourney from zero to control", "ChatGPT Images & Nano Banana", "Google Veo 3 — commercial video", "Video neural networks 2.0" — Plus plan and above, or $9 one-off, or 1 000 experience points. The course "Photo and Video" is included in the Plus plan.
Where to start today
One action. Take your last completed job and calculate one figure for it: how many generations you made before getting the result you delivered.
Not how much money you spent — how many attempts there were.
Usually it turns out two or three were successful and the rest went on feeling out the task. That is your real profit reserve — and it lies not in raising the price but in stating things more precisely and running drafts on cheap models.
The full table of modes
To cost a job you need all the figures at once.
| Mode | Models | Price per generation |
|---|---|---|
| Images | 30 | $0.12 – 0.25 |
| Text to video | 51 | $0.22 – 11.95 |
| Photo → video | 84 | $0.17 – 11.95 |
| Photo editor | 34 | $0.12 – 0.50 |
| Upscale | 6 | $0.12 – 0.40 |
| Background removal | 1 | $0.12 |
| Face Swap | 1 | $0.12 |
| Product Shot | 2 | $0.12 |
| Lip Sync | 13 | $0.12 – 1.50 |
| Video editor | 39 | $0.12 – 11.95 |
| File → text | 4 | $0.15 |
| 3D models | 8 | $0.36 – 0.72 |
The spread in video is fiftyfold. That is where quoting mistakes get made: a draft on a top model eats the margin faster than anything else.
Prices visible before you press the button
The Gen AI section — 338 models on a single balance. Top up from $5, debited on usage, the price of a specific model is written beside it:
| What we are doing | Price per generation |
|---|---|
| An image | $0.12 – $0.25 |
| Video from text | $0.22 – $11.95 |
| Animating a photo | $0.17 – $11.95 |
| Lip sync | $0.12 – $1.50 |
| Music and sound | $0.12 – $1.80 |
| Transcribing a file | $0.15 |
| Upscale | $0.12 – $0.40 |
The spread in video is easily explained: a draft at 480p costs cents, a final run at high resolution costs dollars. Hence the working order: a cheap draft first, then the expensive final on an approved frame.
A ready pipeline instead of assembling by hand. Describe the clip in plain words in Cerebrum — it picks the models and makes the script, frames, animation and narration. The estimate is shown before launch and money for anything not delivered returns to the balance automatically. If you work in Claude, the AI SKILLS connector for Claude (Full plan) does the same, including the assembly.
What that changes in conversation with a client. You name a price knowing your own cost to the cent. That is a rare advantage: usually a contractor works by feel.
A real benchmark for a finished clip
A separate figure, more useful than ranges by operation.
In a chain where the script, the photograph, the animation, the narration and the assembly run automatically, the cost of one clip is $2.5–3. That includes the video model and the assembly rather than only generating a frame.
Compare with the market: a short clip sells for $18–45, turnkey AI video for $245–550.
What to take on the platform for this task
- Video from idea to result — a pipeline skill: the estimate is shown before every paid step; files from Basic
- Product showcase — a skill: a cinematic 5–15 second product ad
- Ad creatives in a set — a skill: 6–9 static creatives in one pass
- Google Veo 3 — commercial video — an express course, $9, Plus plan and above
- Video neural networks 2.0 — "East vs West" — an express course on Kling 3.0 and Seedance 2.0, $9, Plus plan and above
What to read next
LLM pricing and token economics — how this is calculated systematically.
The AI freelancer price list — the first half of the equation.
AI video generation — what exactly you are generating.
What it costs to try — how paying on usage works.
AI video models in 2026 — which model for which task and what it costs.