NodeValid

Published video-model weights, licences and hardware notes

CogVideoX calls 161 frames five or ten seconds

CogVideoX1.5-5B states up to 161 frames, described as five or ten seconds. Two durations for one frame count means two frame rates are in play, and neither is written down. As of 2026-09-12.

CogVideoX in the length column, as publishedGiving a frame count and two possible durations tells a reader the output can be played at more than one rate, without saying which rates those are.Length as publishedUp to 161 frames on 1.5-5Bframes, seconds or bothDescribed as 5 or 10 secondsOutput size1360x768 on 1.5-5Bpixels per frame720x480 on the 5BMemoryFrom 4GB in FP16 on the 2Bwhat holding them costFrom 5GB on the 5B, from10GB on 1.5-5BOne clip, as this release describes itTwo durations, one frame count
Fig. 1 One frame count offered as two clip lengths, with the rates left to be inferred.
CogVideoX in the length column, with the columns that qualify it. Read from github.com/THUDM on 2026-09-12.
ColumnWhat the pages state
Length, in the vendor's own wordsUp to 161 frames on 1.5-5B
The detail published beside itDescribed as 5 or 10 seconds
Resolution, from the same pages1360x768 on 1.5-5B
Hardware, from the same pagesFrom 4GB in FP16 on the 2B

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.

1Five or ten is an unusual way to publish a length

161 frames at 32 FPS is about five seconds; at 16 FPS it is about ten. Offering both without naming either says the rate is a playback choice rather than a property of the generation.

That is technically reasonable and it makes the value hard to compare. Another release here states 88 frames, six seconds and 15 FPS, which admits no ambiguity at all.

2The frames belong to a named release

The 161 figure is stated for CogVideoX1.5-5B, which is also the release with the 1360x768 output and the 10GB memory floor. Every value on that row names the same model.

The 5B release, at 720x480, has no separate length figure. So the family publishes one length for one of its two documented sizes.

3Why the register keeps the phrasing

Recording five seconds would drop the alternative; recording ten would drop the other. Picking a rate to resolve it would put a number in the register that the repository avoided publishing.

The value goes in as written, with the frame count first, because the frame count is the part that is unambiguous.

  • Length and resolution
    CogVideoX1.5-5B states 1360x768 and up to 161 frames, described as 5 or 10 secondsper modelTHUDM, CogVideo repository / recorded 2026-09-12
  • Licence
    Split: CogVideoX-2B is Apache 2.0 while CogVideoX-5B carries a separate CogVideoX licencediffers by model size inside one familyTHUDM, CogVideo repository / recorded 2026-09-12

4Sources

Lengths above are in the unit the vendor chose, from github.com/THUDM on 2026-09-12, with no conversion applied. The full entry is at this family's page; frames against seconds is on length. Nearby: No length recorded, Frames, seconds, rate.