A flat retry allowance treats a locked-off product shot and a complex moving character scene as equally risky. They are not necessarily equally difficult. Before generating anything, assign each storyboard beat a purpose, shot length, and planning risk tier. These tiers are internal estimates, not claims about any provider.
A sample storyboard and credit envelope
Assume twelve six-second shots, 72 seconds of final approved footage, and a hypothetical rate of 4 credits per generated second. Each attempt would use 6 × 4 = 24 credits. Set a different attempt ceiling for each beat:
| Shot group | Shots | Attempts per accepted shot | Planned attempts | Credits |
|---|---|---|---|---|
| Simple inserts | 4 | 1.5 | 6 | 144 |
| Medium motion | 5 | 2.5 | 12.5 | 300 |
| High-constraint scenes | 3 | 4.0 | 12 | 288 |
| Total, expectation | 12 | — | 30.5 | 732 |
Fractional attempts appear here because we are estimating the expected workload across a group, not claiming half an actual generation. For procurement, round each group's attempt count upward to whole generations, producing a slightly more conservative cap. In this example, the whole-number group budget is 6 + 13 + 12 = 31 attempts, or 744 credits.
Define what makes a shot difficult
The phrase “high-constraint” needs a reason. It might mean precise product text must be legible, a person must maintain consistent clothing, or a camera move must match an existing take. The label should come from the brief and early tests, not an invented universal “success rate.” A low-risk scene may become high-risk once the client specifies a hard-to-reproduce detail.
For each shot, capture these five fields in the storyboard: brief requirement, planned duration, source/reference rights, maximum attempts before review, and acceptance condition. “Looks cinematic” is a weak acceptance test. “The label is readable for at least two seconds” is stronger.
Do not spend the entire reserve immediately
Use a two-stage release. In stage one, generate one trial per shot, check continuity, and identify which constraints are failing. Stage two spends the retry reserve only where needed. This avoids exhausting credits on a hero shot before discovering that the overall style is wrong.
When several shots fail for the same reason, stop. Rewrite the shared prompt or change the creative approach rather than multiplying the same failure. A stop-loss is useful in creative production: after a defined number of unsuccessful attempts, return the shot to the editor for a decision.
Keep a usable acceptance log
A tiny spreadsheet works: shot ID, attempt ID, generation length, credits charged, keeper yes/no, reason, approved usable duration. At the end of a test project, compute total credits / total approved seconds. Do not treat a hand-picked collection of successful generations as a complete benchmark; the failed attempts belong in the denominator as costs.
Next step: enter your 72-second project and measured average attempts into the calculator. If your actual shots vary in length, estimate each length group separately and add the results.
Check the source and keep your inputs
Our examples show a decision-making method, not a live price quotation. Before purchasing credits, verify the provider’s billing method and current plan rules in its official credit documentation or the equivalent official page for the tool you use. Keep your calculations reproducible with our shot-by-shot CSV worksheet. See the methods and sources page for limitations and a full cost ledger.