AI Video Budgets for Agencies: Calculate Cost per Approved Clip
Clients do not buy prompts or generation attempts. They buy approved video clips.
That makes cost per approved clip a more useful budgeting unit than cost per generation. An output may be rejected because it misses the brief, contains technical problems, or requires too much correction.
For agencies and production teams using tools such as Google Flow or Vertex AI, the budget needs to connect generation expense with a client-ready deliverable.
In its October 2025 announcement, Google described Veo 3.1 as available through Flow, the Gemini API, Vertex AI, and the Gemini app. Before using any provider rate in a quote, check the current official pricing and usage terms.
Pricing note: Flow credit allocations and credit costs, Vertex AI rates, subscription prices, supported configurations, regional terms, taxes, and promotions can change. This guide deliberately does not quote current provider rates. Verify Google Flow credit rules and Vertex AI pricing before quoting.
Why approved clips are a useful planning unit
A generation is a production input. An approved clip is the client-facing output.
The gap between them can include rejected outputs, additional attempts, model and resolution choices, audio requirements, prompting and creative-direction time, editing, review cycles, storage, transfer, music, and voice costs.
A subscription price or prompt count may be easy to report, but neither tells you what one accepted deliverable actually costs. Start with the deliverable and work backward.
The approved-clip formula
Expected generations per approved clip = 1 ÷ approval rate
Direct tool cost per approved clip
= cost per generation × expected generations per approved clip
Total production cost per approved clip
= direct tool cost + labor + post-production
+ storage and transfer + music and voice
+ review and delivery
An approval rate of 75%, for example, implies approximately 1.33 expected generations per approved clip. That is a planning assumption, not a promise about a particular brief. Record results by project and workflow so the estimate improves over time.
Budget worksheet
| Input | What to record | Example treatment |
|---|---|---|
| Approved clips | Number required by the client | Project requirement |
| Expected generations per approved clip | Historical approval data | Use a measured rate or label the assumption |
| Provider usage | Generations, credits, or API usage | Keep Flow and Vertex AI records separate |
| Allocated provider cost | Metered spend or subscription allocation | Document the allocation method |
| Labor hours | Prompting, selection, production, and management | Multiply by internal cost rate |
| Post-production | Editing, compositing, upscaling, and formatting | Add as a project line item |
| Review work | Internal review, client review, and revisions | Estimate separately from generation |
| Target margin | Required contribution margin | Use the pricing formula below |
There is no single project-independent rate for labor, review, storage, or post-production. Use your own time records, invoices, and project data. If a value is not measured yet, mark it as an input rather than presenting it as a provider fact.
An AI video cost calculator for project margin
The spreadsheet needs these editable inputs:
A = approved clips required
G = expected generations per approved clip
C = cost per generation
L = labor cost for the project
P = post-production cost
R = review and delivery cost
O = other production costs
M = target contribution margin as a decimal
Tool cost = A × G × C
Total project cost = tool cost + L + P + R + O
Client price = Total project cost ÷ (1 - M)
Contribution margin = (Client price - Total project cost) ÷ Client price
Consider a hypothetical 10-clip project. The $0.40 generation allocation below is invented for this example, not a current Google price:
- 10 approved clips required;
- 2.5 expected generations per approved clip;
- hypothetical generation allocation: $0.40;
- labor: $420;
- post-production: $180;
- review and delivery: $100;
- other production costs: $50;
- target contribution margin: 40%.
Tool cost = 10 × 2.5 × $0.40 = $10
Total project cost = $10 + $420 + $180 + $100 + $50 = $760
Client price = $760 ÷ (1 - 0.40) = $1,266.67
Contribution margin = ($1,266.67 - $760) ÷ $1,266.67 = 40%
In this example, tool cost is $1 per approved clip, total project cost is $76 per approved clip, and the client price is about $126.67 per approved clip. Replace the hypothetical inputs with your project records before quoting.
Billing units are not deliverables
For a Flow-based project, record the subscription or credit expense you allocate to that project, using a documented allocation rule. For an API-based project, record the actual billed usage for that project. Avoid charging the same expense twice. Spendly’s guide on how to allocate AI costs by project can help structure that allocation process.
Before entering a provider rate, check its billing unit. A rate per second, per request, per generated output, or per credit cannot be inserted into the calculator unchanged. Convert the billed units into your actual project cost first; then divide by approved clips.
If a request creates more than one candidate output, count the outputs consistently when measuring approval rate. One approved deliverable may also combine several source clips. In that case, budget those source clips as separate production inputs.
Do not infer a cash conversion between Flow credits and API usage. Compare the resulting project cost, accepted deliverables, and production time.
A practical model comparison
Compare candidate models or modes using the same brief and delivery requirements. The better choice depends on the full production path, not just the provider rate.
| Question | Try a lower-cost workflow when… | Test an alternative workflow when… |
|---|---|---|
| Approval rate | Historical approval is strong | The brief is demanding or approval is uncertain |
| Review burden | Producers can select usable outputs quickly | Review and correction are consuming substantial time |
| Turnaround | The project needs rapid exploration or iteration | A final shot must meet a high bar before delivery |
| Resolution | The deliverable fits the supported lower resolution | The brief requires a higher-resolution output |
| Audio | Audio is added later or is not required | Generated audio is part of the approved deliverable |
| Total cost | Lower tool cost also reduces project cost | A higher tool cost may reduce failed attempts and labor |
No tier wins every brief. A cheaper generation may need more selection or revision work, making the final deliverable more expensive.
From the worksheet to a client quote
The calculator covers per-clip production and project-level costs. Check that creative development, setup, producer time, client meetings, and delivery administration are included once—not in several overlapping line items.
For an uncertain brief, quote a documented range or revision allowance. Define the escalation point before work starts: who approves additional attempts, how the client is notified, and which changes fall outside the agreed scope. Adapt the margin calculation to your agency’s taxes, overhead allocation, and commercial terms.
Project reporting in Spendly
Spendly offers provider, model, employee, project, budget, forecast, success-rate, failure, cost-per-result, and project-economics views. These views help connect AI usage with the cost of client deliverables. This budgeting worksheet does not imply a native integration with Google Flow.
For an AI video workflow, useful project fields include provider and model, project or client, generations and failed generations, resolution and audio mode, approved and rejected clips, labor and review time, allocated subscription or metered provider cost, revision rounds, and final deliverables.
Review the data weekly or at project milestones. Use AI cost alerts to define thresholds for unexpected spend, rising failure rates, or projects drifting beyond their planned allowance.
Spendly can help teams view AI spend by provider, model, employee, and project, set budgets and alerts, compare alternatives, review failed-output costs, and connect usage with project economics.
When to update the worksheet
AI video pricing has a high expiration risk. Recheck the worksheet when Flow changes credit allocations or rollover rules, Vertex AI changes model rates or supported resolutions, audio options change, a new Veo model replaces a documented workflow, regional terms or promotions change, or actual approval and review data diverges from assumptions.
A dated source check is part of the budget, not an administrative extra. Verify current Google pricing before quoting.
Try it on your next brief
Pick one upcoming video project. Fill in the worksheet before generation starts, name the person who accepts each deliverable, and set a review date after the first batch. For a broader planning process, use these principles to set an AI budget before committing to recurring tools or workflows. Compare the estimate with actual costs before committing the remaining budget.
FAQ
What is the cost per approved AI video clip?
There is no universal figure. Calculate expected generations per approved clip, multiply by the applicable tool cost, then add labor, post-production, storage, review, delivery, and other project costs.
Can I use this as an AI video cost calculator?
Yes. Put the worksheet variables into a spreadsheet: approved clips, expected generations, cost per generation, labor, post-production, review, other costs, and target margin. Keep provider rates in editable cells and label hypothetical values clearly.
Are Google Flow credits directly comparable with Vertex AI API prices?
No direct conversion is established here. Keep the ledgers separate and compare workflows through approved clips, generations, spend, labor, review time, and total production cost.
How should an agency compare AI video models?
Test the same brief across candidate models. Record the approval rate, review time, delivery requirements, and total production cost. Choose from measured project results, not a claim that one tier is always better.
How should rejected generations be accounted for?
Count rejected candidate outputs in the generation total, but not in approved deliverables. Their tool cost remains part of the project cost. Record the rejection reason so the next budget can distinguish technical failures from creative revisions.
Which production and review costs belong in the budget?
Include paid production time, editing, post-production, storage, review, and delivery where they apply. Assign each expense once: a cost already included in the labor total must not be added again under review or project management.
Official sources and references
Spendly Editorial
Practical guidance for teams that need clear ownership and control of AI spending.