NodeValid

Published video-model weights, licences and hardware notes

Step-Video-T2V caps at 204 frames, no rate

The repository states videos up to 204 frames. No frame rate appears anywhere on these pages, so the ceiling is a count rather than a duration and cannot be converted into one. As of 2026-09-22.

Step-Video-T2V in the length column, as publishedA frame ceiling without a rate bounds the work the model will do and leaves the length a viewer would perceive unstated.Length as publishedUp to 204 framesframes, seconds or bothFrames, with no frame rateattachedOutput size768px768px and 544px992pxpixels per frameStated as the memory table'sconfigurationsMemory78.55 GB peak at 768px768px204fwhat holding them cost77.64 GB and 72.48 GB at twosmaller settingsOne clip, as this release describes itCounted, never timed
Fig. 1 A ceiling in frames, sitting on a memory table that uses the same unit.
Step-Video-T2V in the length column, with the columns that qualify it. Read from github.com/stepfun-ai on 2026-09-22.
ColumnWhat the pages state
Length, in the vendor's own wordsUp to 204 frames
The detail published beside itFrames, with no frame rate attached
Resolution, from the same pages768px768px and 544px992px
Hardware, from the same pages78.55 GB peak at 768px768px204f

Inclusion rule. The value this page is about, the detail the vendor attached to it, and the two columns that change how it should be read. A column the vendor left empty keeps its row and says so. Order. This column first, then its detail, then the two columns that qualify it.

1The unit is consistent across the page

The memory table also labels its configurations in frames, 204f and 136f, so frames are the unit this repository thinks in. The omission of a rate is coherent rather than careless.

It does mean the register cannot place this release beside the ones that publish seconds. 204 frames is somewhere between seven and thirteen seconds depending on a value nobody wrote down.

2What the ceiling costs

78.55 GB peak at 768px768px with 204 frames, against 72.48 GB at 544px992px with 136. Going from 136 to 204 frames is part of a six gigabyte swing that also changes the frame shape.

So the longest published clip is affordable on the same class of machine as the shortest. For a thirty billion parameter model the frames are a small share of the total.

3Why the register does not assume 24 FPS

Because a plausible default is still an invention. Another release in this register uses 16 FPS and a third uses 15, which is enough to show that no default is safe.

The value stays in frames. Anyone needing seconds has to establish the rate from the code rather than from this entry.

4Sources

Lengths above are in the unit the vendor chose, from github.com/stepfun-ai on 2026-09-22, with no conversion applied. The full entry is at this family's page; frames against seconds is on length. Nearby: Length as a code rule, Four seconds, admitted.