RenderDrama

What a second of AI video costs, channel by channel

What three sellers charge for Seedance

Three channels here publish a figure against a Seedance model and they agree to within half a cent: nine point seven cents, ten, and ten point two five. No other family in this register is priced that consistently across sellers. As of 2026-09-22.

A tight spread, and what is really being chosenThree sellers land within half a cent of each other, which is the tightest agreement any family reaches here. Two of the three figures carry a condition: one is a plan price on an annual commitment, one sits under a page-wide clause saying prices may vary.Cheapest by figureFirmest by conditionThe rowNine point seven cents a secondTen point two five cents a secondWhat it needsAn annual plan commitmentNothingVersion namedA 2.5 release at 720pA 2.5 releaseGap to the otherHalf a cent cheaperHalf a cent dearerWho should pick itA production already on that planAnyone comparing month to monthHalf a cent of spread moves the decision to version and conditions
Fig. 1 Ranked by what a buyer can rely on, the order reverses: the dearest row is the only one needing neither a commitment nor an assumption.
Every channel here that publishes a figure against a Seedance model, and its conditions. Recorded 2026-09-22.
ChannelPublished figureVersion and condition
DreaminaNine point seven cents a secondSeedance 2.5 at 720p, on one plan billed annually
WaveSpeed AITen cents a secondSeedance 2.0 Fast, under a page-wide variability clause
RunwareTen point two five cents a secondSeedance 2.5, unconditional on the page

Inclusion rule. Channels in this register publishing a figure against a Seedance model on a public page. Conditions are named because two of the three figures carry one. Order. Ascending by published figure.

1The production behind the price

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

The Seedance 2.0 tiers render segments up to 15 seconds; use a Seedance 2.5 or Wan tier for 30-second takes.

SceneMixer, Seedance 2.0 Fast — “Does Fast render 30-second takes?” (read 2026-10-13)

In the editor this tier is greyed out for such a segment; pick a Seedance 2.5 or Wan tier.

SceneMixer, Seedance 2.0 (standard tier) — “Can a segment longer than 15 seconds use this tier?” (read 2026-10-13)

2Where this family sits against the others

Cheapest published second by model familyThe lowest second published for each family across every channel here, Seedance marked. A conditional figure is excluded, so the floor shown is one a buyer reaches without negotiating or committing. Recorded 2026-09-22.Cheapest published second by model familySeedance$0.097 a second$0.097 a secondLTX-2.5$0.09 a second$0.09 a secondKling$0.07 a second$0.07 a secondVeo$0.05 a second$0.05 a secondMiniMax H3$0.046 a second$0.046 a secondWan$0.01 a second$0.01 a second
Fig. 2 The lowest second published for each family across every channel here, Seedance marked. A conditional figure is excluded, so the floor shown is one a buyer reaches without negotiating or committing. Recorded 2026-09-22.
The cheapest unconditional second published for each model family here, with Seedance marked. Recorded 2026-09-22.
Model familyCheapest published second
Kling$0.07 a second
LTX-2.5$0.09 a second
MiniMax H3$0.046 a second
Seedance$0.097 a second
Veo$0.05 a second
Wan$0.01 a second

Inclusion rule. Model families with a figure from more than one channel in this register, taking the lowest figure any of those channels publishes without a discount or a commitment attached. Order. Alphabetical by model family.

3Agreement this tight usually means a fixed wholesale rate

Half a cent of spread across three separate sellers is not competition; it is three shops passing through the same input cost with thin and similar margins. Where a family behaves like that, choosing a channel on price is choosing between rounding errors.

It also makes the family a good control. Anywhere else in this register that a spread opens up, the question is what the sellers are doing differently, and this page shows what the baseline looks like when they are not doing anything differently at all.

4Two of the three figures are conditional, and differently so

The cheapest row is a plan price on an annual commitment, so a monthly buyer does not reach it. The middle row sits under a clause saying prices may vary with resolution and parameters. Only the dearest row is unconditional on its own page.

Ranked by what a buyer can actually rely on, the order reverses. The row a comparison would put last is the only one that does not need a commitment or an assumption to be true.

5Versions differ even where the price does not

Two rows name a 2.5 release and one names a 2.0 fast variant. A fast variant at the same price as a full release is a worse purchase on quality and a better one on latency, and the rate card cannot express either.

So the useful reading of a tight spread is that the decision has moved elsewhere: to version, to conditions, to what else the channel publishes. That is a better position for a buyer than a wide spread with nothing to explain it.

6Sources

Every figure is read from the page linked on that channel's own entry, checked 2026-09-22. What counts as a published rate is on the method page; the channels in full are on the ledger. Other families: LTX-2.5, Wan. All of them: one model, several sellers.