Costing an episode before you generate it

Start from delivered minutes, not from the rate. Multiply by your own attempt ratio, add the joins your clip ceiling forces, then apply a published rate. The rate is the last step and the least uncertain one. As of 2026-09-12.

Inclusion rule. Rates that appear on the ledger, spanning its published range. The arithmetic is ninety seconds of delivered picture; nothing here is a quote and no vendor is named on this row. Order. Ascending by rate per second.

1Start from what gets delivered

A short-drama episode is typically one to three minutes of finished footage. That is the only number in the whole calculation that comes from the creative brief rather than from a vendor, and everything else scales off it. Write it down first and resist adjusting it to make the arithmetic comfortable.

For a season, multiply by episodes and then stop. Season-level numbers feel motivating and hide the per-episode decisions where money is actually spent or saved.

2Measure your attempt ratio on one episode

Generate one episode with normal standards. Count every second you generated, including the discards, and divide by the seconds you delivered. That number is your attempt ratio and it is the most important figure in your budget.

Teams new to this routinely assume something near one and a half. Measured ratios in the field vary far more widely than that, in both directions, depending on how tightly the material is specified before generation. A well-specified shot list with locked character references is the cheapest cost control available, because it acts directly on this multiplier.

Re-measure it after any change to the model, the prompt template, or the reference set. It is not a constant.

3Count the joins your clip ceiling forces

Divide the delivered duration by the service's maximum single generation and round up. That is the minimum number of clips, and one less than that is the number of joins. Each join carries a risk of visible discontinuity and therefore a chance of regeneration.

Add a modest allowance per join rather than a flat percentage on the whole project. Joins are where the cost is, and spreading the allowance evenly hides which decisions would reduce it.

4Add the steps that are priced separately

Some services charge for compositing segments into a finished piece; some leave assembly to an editor, which is a different budget line rather than a free one. Either way the step exists, and a budget that omits it is describing clips rather than an episode.

Sound, music licensing, subtitles and delivery encoding sit outside every generation rate in this market. They are usually small next to generation and they are never zero.

5Apply the rate last, and sanity-check against a minute

Only now bring in the published per-second figure, at the resolution you actually need. Multiplying it by sixty gives a cost per minute that is easy to hold in your head and easy to challenge in a meeting.

If the resulting episode number looks impossible, the error is almost never in the rate. It is in the attempt ratio or in an unrealistic delivered duration, and both are yours to fix.

6Where the savings actually are

Ranked by how much they move the total: the attempt ratio first, by a wide margin; then the resolution tier; then the clip ceiling; then the rate itself. Most teams work that list backwards, spending a week comparing rates that differ by cents and no time at all on the multiplier that differs by whole numbers.

Acting on the attempt ratio is unglamorous work: lock the character references before generating, write the shot down before prompting it, decide the camera move rather than discovering it, and stop regenerating for differences an audience will not notice. None of that is a purchasing decision, which is probably why it gets less attention than the price list.

The resolution decision is a single choice per project and worth making deliberately once rather than defaulting. The clip ceiling is not something you can change, but knowing it before the shot list is written lets the shot list respect it, which removes joins rather than paying for them.

7A worked shape, with the arithmetic left visible

Take a two-minute episode: 120 delivered seconds. With a measured attempt ratio of two, that is 240 generated seconds. At a 480p-class rate of roughly six cents, generation is around fifteen dollars. A thirty-second clip ceiling gives four clips and three joins; allow a little regeneration for those and it is perhaps eighteen. Add assembly, sound and delivery and the episode has a defensible number.

The figure is not the point. The point is that every input is either from your own measurement or from a page you can link, so when somebody challenges the total you can show which input they disagree with instead of arguing about the total.

8Common mistakes worth naming

Budgeting from the cheapest tier and generating at the highest. Comparing a credit price against a currency price without converting. Treating a vendor's demo reel as evidence about your attempt ratio. Assuming the clip ceiling will not matter because your shots are short, then discovering the edit wants a longer take.

All four are avoidable with a page of arithmetic done before anything is generated, which is the entire argument for doing it in that order.

9Where the sourced entries live

This page is about how the pricing works. What any particular channel charges is a separate matter, kept in the ledger with the page each figure came from and the date it was read.

• SceneMixer — publishes a per-second rate per named model tier
• Vidu — publishes rates by resolution, with an off-peak condition
• Runway — publishes credits per second with the credit price alongside

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: Costing a season, Token billing.