AI Video Cost Calculator
Published per-second pricing tells you what a render costs, not what a finished shot costs. Retakes are the multiplier nobody prices in. Put your real numbers in and see the difference.
Ambient b-roll often lands in 1–2. Anything with characters, precise motion or lip-sync realistically runs 3–5.
Presets are published ranges, not quotes. Check your own invoice before committing a budget — see our cost breakdown and Luma API pricing.
Cost per finished shot
$1.68
- Delivered footage
- 100s
- Seconds actually rendered
- 300s
- Total project cost
- $33.60
Drafting first saves $38.40 (53%) versus rendering every take at full price ($72.00).
The multiplier most pricing pages ignore is takes per keeper. At 3×, you pay for 300s of rendering to deliver 100s of video.
Frequently asked questions
▸Why is cost per second the wrong number?
Because it prices rendering, not delivery. A model at $0.10/sec that needs five attempts to nail a camera move costs more per finished shot than a $0.15/sec model that gets there in two. Published pricing pages advertise the first number and never mention the second, which is why a cheaper-on-paper model regularly loses on a real production budget.
▸How many takes per keeper should I assume?
Start at 1–2 for ambient b-roll and simple motion, and 3–5 for anything involving characters, precise camera timing, or lip-sync. Those are our own operating averages across production work, not vendor-published figures. Track your own rate for a week and replace the assumption — it is the single biggest lever in the whole calculation.
▸What does 'draft first' mean here?
Render throwaway attempts at low resolution or short duration to confirm the take, then pay full price only for the keeper. Because the discarded takes bill at a fraction of a full render, the saving scales with your retake rate: the more attempts a shot needs, the more the staged workflow beats rendering everything at final quality.
▸Are the preset prices accurate for my account?
Treat them as published ranges to sanity-check against, not quotes. Vendor pricing in this category changes faster than any static table, and resolution, dynamic range and native audio each move the rate independently. Enter your own per-second figure from your invoice for a number you can actually budget against.
Where these numbers come from
Preset rates are drawn from published API pricing trackers, which disagree with each other by a meaningful margin on mid-tier models — itself a useful signal about how fast this market moves. The retake ranges are our own production averages, and we say so rather than dressing them up as vendor data. For the full reasoning, read the cost-per-finished-shot breakdown.