Clip length, and what it changes about a plan
Clip length is how long a single generation runs for. It decides whether a shot arrives whole or has to be assembled, and published values in this register span from four seconds to a stated minute. As of 2026-09-22.
| Aspect | As length grows |
|---|---|
| Frames held at once | Rises |
| Continuity work | Falls |
| Memory needed | Rises |
| Failure cost per run | Rises |
Inclusion rule. Aspects of a pipeline that move with clip length. Rows describe regimes, so no vendor figure appears. Order. From the mechanical consequences to the operational one.
1The failure cost is the underrated one
A failed four-second generation wastes a few minutes. A failed sixty-second generation wastes far more of a card's time, and generative video fails often enough for that to matter in a capacity plan.
None of the releases in this register publish a failure rate, so that cost has to be estimated by whoever runs the pipeline.
2Long output is where the columns interact
Length, resolution and memory compound: more frames at more pixels held at once. The release in this register making the longest claim publishes neither a frame size nor a memory figure.
So the longest stated clip length is also the least costed. The entry records the claim and the three gaps around it.
3The span the register covers
Published lengths here run from four seconds, written as a limitation, to a stated minute written as supported. The column mixes releases from different years without adjustment.
Two families publish no length at all and one describes its output with an adjective. So the column is well filled at both extremes and has three cells that cannot be planned against.
A reading note, not an entry: no vendor value appears on this page. Where the register records this term for a particular release, it is on the length column and on that family's own page. Nearby terms: Frame interpolation, Temporal drift.