What a ten-episode season actually costs
Ten ninety-second episodes is nine hundred seconds of delivered picture. At published rates that is from about thirty to several hundred dollars of generation, and it is nowhere near what the season costs. The gap is five items, none of which appear on a rate card. As of 2026-09-12.
| Rate per second | Generation only | With one retake in three | Per episode |
|---|---|---|---|
| $0.035 | $31.50 | $42.00 | $4.20 |
| $0.06 | $54.00 | $72.00 | $7.20 |
| $0.12 | $108.00 | $144.00 | $14.40 |
| $0.42 | $378.00 | $504.00 | $50.40 |
Inclusion rule. Rates spanning the published range on the ledger. The arithmetic is delivered picture only; nothing here is a quote and no vendor is named on a row. Order. Ascending by rate per second.
1The five costs that are not on the rate card
Reference art for the cast and locations, built once. Script breakdown, which is hours of a person. Discarded attempts, which are real generations at the same rate. Assembly, which some channels price and most do not. Storage and delivery, small but recurring.
Only the third of those scales with the published rate. The rest scale with the number of episodes and with how organised the production is, which is why two teams on the same channel can differ by more than the gap between the cheapest and dearest rows above.
Reference art is the item most often left out of a first budget and the one that pays back most clearly, because every episode after the first draws on it. Assembly is the item most often underestimated, since ninety seconds of finished picture is made of many more pieces than a shot list suggests.
2The attempt ratio is the number nobody publishes
No vendor states how often a generation is kept. Productions measure it for themselves, and the figure moves with the shot: dialogue close-ups are cheap to get right, crowd and action shots are not.
A season budgeted at one attempt per shot is budgeted wrong. One retake in three is a conservative working assumption and is what the third column above uses.
Because the ratio is measured rather than published, it is also the one number a production can actually improve. Tighter reference art and a shot list written for what the channel does well both move it, and both are decisions made before a single second is generated.
3Where the arithmetic goes wrong most often
Confusing a segment with a shot. Generation is billed per second of returned video, so a season planned as four hundred shots on a model that cuts inside one call may be fifteen dozen calls rather than four hundred.
The second most common error is comparing headline rates across channels that deliver at different resolutions. The rate that matters is the one at the resolution the season ships at.
The third is forgetting that a rate quoted in credits is not a rate until the credit price is known, and that some channels sell credits in bundles that expire. A season planned across three months on a monthly bundle can pay for capacity it never uses.
4What this means against live action
The comparison people want is with a filmed production, and it is not a comparison this register can make: filmed budgets are dominated by people and time, neither of which has a published per-second rate.
What can be said is narrower and more useful: the generation line is small enough that it is rarely the constraint. The constraint is the four items around it.
A team that has already made one season will spend less on the second, and almost none of that saving comes from a cheaper rate. It comes from reference art that already exists, a breakdown method that is already settled, and a lower attempt ratio because the shots that fail are known in advance.
Written to explain, not to record. No sentence here is attributed to a seller, and none of it should be quoted as a fact about one. The sourced material is on the ledger. Related: Token billing, Resale and source.
5Inside a generated second
Price aside, what goes into the seconds behind “Costing a season” is described in SceneMixer's guides:
That is SceneMixer's core design: characters live in a persistent asset library with locked reference art, and every segment that features a character is generated against those references — same face in episode 1 and episode 50.
SceneMixer, The AI Micro Drama Production Pipeline — “Can AI keep the same actors across 50 episodes?” (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)