Video generation settles on per-second billing, and the numbers are now small
MiniMax H3 bills at $0.13 per second for 2K output and $0.09 per second at 768p. A fifteen-second 2K clip with native stereo audio therefore costs under two dollars — and a per-second meter, rather than per-clip or per-seat, is what makes video a line item instead of a project.
The pricing unit is the story, not the price. Per-second billing turns generated video into something a spreadsheet can hold: a 30-second spot has a cost you can compute before you commission it, and so does the fourth revision of it.
The arithmetic is stark against the alternative. $0.13 per second at 2K puts a fifteen-second clip at $1.95. At 768p it is $1.35. Those are numbers that compete with stock footage licensing, not with production, and they include natively generated stereo audio rather than a separate scoring pass.
The second-order effect is on iteration. When a clip costs two dollars, the constraint on trying twenty variants is the reviewer's attention, not the budget. That inverts the economics of the edit: generation becomes cheap and selection becomes the expensive, human, unautomated step.
Which is roughly where language models landed two years ago — and it produced the same result there. The bottleneck moves to whoever has to decide which output was any good.
AIbase — Multimodal Generation Upgraded: 2K HD Audio and Video → · Pinggy — Best Video Generation AI Models in 2026 →