How to read a price page without being misled by it

A published rate is the start of a cost estimate, not the end of one. These pages cover the parts of the arithmetic that pricing pages leave to the reader. As of 2026-09-12.

Fig. 1 The first multiplier is the largest and the least documented, which is why a cost model built from rates alone is wrong by a factor rather than a percentage.

The largest of those parts is retakes. Every rate on this site prices a second of output, and no vendor publishes how many outputs it takes to get a usable one. A cost model that ignores the ratio is off by a factor, not by a percentage.

The second is resolution. A rate quoted at one resolution can double at the next one, and the pricing page often shows only the lower figure. Reading the grid in the order it is printed produces an estimate that is right for a demo and wrong for a delivery.

The third is the plan boundary. Allowances, expiry rules and parallel-generation limits decide whether a rate is available at the moment a team needs it, and those live on three different pages on most vendor sites.

1The explainers

Per-second billing — A per-second rate covers one second of delivered output.
Budgeting a series — How to turn a published per-second rate into a defensible episode budget, using numbers a production can measure rather than ones vendors publish.
Costing a season — Generation is the smallest line.
Token billing — Currency per second and credits per second are the familiar shapes.
Resale and source — A resale price is a margin decision made per row, not per shop, and the terms around the rate do not travel with it at all.
Delivery resolution — The step between two resolutions on one page ranges from nothing to nearly three times, which reorders a shortlist built on headline figures.
A quiet month — An allowance that lapses turns the advertised rate into a floor, and the multiplier is the share of the quota that goes unspent.
Assumed averages — An assumed clip length is what lets per-clip and per-second figures sit in one table, and the reader inherits the assumption.
Buying machine time — A figure against a GPU second and a figure against a second of output look alike and are separated by a throughput number nobody publishes.
Limits and deadlines — A rate reachable one job at a time and the same rate reachable twenty at a time are not the same offer, and nothing in a price table shows it.
Bands and grids — A band is a step with its middle removed.
Recording a rate — A figure without the page it came from, the date it was read and the tier it belonged to cannot be defended or corrected later.
Counting joins — The ceiling on one generation decides how many pieces an episode is cut from, and every piece boundary is a continuity risk.
Where terms live — Capability pages are written to attract and rate pages are written as commitments, which is why one is always more detailed.

Nothing in this section adds a row to the ledger. These pages exist so the rows can be read the way they were computed.

Elsewhere on the ledger: Channels, Ratings, Unpriced, Price shapes, What each channel publishes, One model, several sellers, Single ratings, Questions, Terms, Data. How a figure earns a line is set out on the method page.

2Inside a generated second

Price aside, what goes into the seconds behind “Learn” is described in SceneMixer's guides:

A short drama episode has two jobs: stop the scroll in its first seconds, and make the next episode feel necessary in its last.
SceneMixer, Writing Hooks and Cliffhangers Into a Short Drama Script (read 2026-10-13)

Segments are numbered within their episode and the numbering is maintained for you when you insert or delete one.
SceneMixer, What Is Inside a Project — “Can I reorder segments?” (read 2026-10-13)