NodeValid

Published video-model weights, licences and hardware notes

Six gigabytes against seventy-eight

FramePack states a 6GB minimum for 1800 frames on a 13B model. Step-Video-T2V states 78.55 GB peak at its largest setting and 72.48 GB at its smallest. The gap is thirteen times and both figures are the vendors' own. As of 2026-09-22.

FramePack and Step-Video-T2V on the same columnOne vendor publishes the smallest card that finishes the job and the other publishes the high-water mark of a large one, which are different claims as well as different numbers.FramePackStep-Video-T2VHardware, as each vendor statesit6GB minimum for 1800 frames at 30fps78.55 GB peak at 768px768px204fThe detail published beside itRTX 30XX, 40XX, 50XX with fp16 and bf1677.64 GB and 72.48 GB at two smallersettingsResolution, from the same pagesNo output size recorded768px768px and 544px992pxA minimum and a peak are not comparable quantities
Fig. 1 A floor and a peak, thirteen times apart, with the output each one buys.
What the repositories actually stateFilled where the repository states something, hollow where it does not. A blank is not an inference.What the repositories actually stateFramePackStep-Video-T2VHardware, as each vendor states…Hardware, as each vendor states itHardware, as each vendor states it — FramePack: 6GB minimum for 1800 frames at 30fpsHardware, as each vendor states it — Step-Video-T2V: 78.55 GB peak at 768px768px204fThe detail published beside itThe detail published beside it — FramePack: RTX 30XX, 40XX, 50XX with fp16 and bf16The detail published beside it — Step-Video-T2V: 77.64 GB and 72.48 GB at two smaller settingsResolution, from the same pagesResolution, from the same pages — FramePack: No output size recordedResolution, from the same pages — Step-Video-T2V: 768px768px and 544px992pxLength, from the same pagesLength, from the same pages — FramePack: 1 minute, 1800 frames at 30fpsLength, from the same pages — Step-Video-T2V: Up to 204 frames
Fig. 2 Filled where the repository states something, hollow where it does not. A blank is not an inference.
FramePack and Step-Video-T2V on the hardware column, read from github.com/lllyasviel and github.com/stepfun-ai. Recorded 2026-09-22.
PointFramePackStep-Video-T2V
Hardware, as each vendor states it6GB minimum for 1800 frames at 30fps78.55 GB peak at 768px768px204f
The detail published beside itRTX 30XX, 40XX, 50XX with fp16 and bf1677.64 GB and 72.48 GB at two smaller settings
Resolution, from the same pagesNo output size recorded768px768px and 544px992px
Length, from the same pages1 minute, 1800 frames at 30fpsUp to 204 frames

Inclusion rule. Two families whose published values in this column lead to opposite decisions. A pair that happens to agree is not listed, because agreement is already visible on the column summary. Order. The column itself first, then its detail, then the columns that qualify it.

1Part of the gap is the question being answered

A minimum says what will work. A peak says what a run consumed on a card that had room. Even for identical weights those two numbers would differ, and here the weights differ too.

So the thirteenfold figure overstates the distance between the models and understates nothing. The useful comparison is between what each vendor chose to publish.

2The rest of the gap is real

Thirty billion parameters against thirteen, with a memory table that barely moves when the setting changes: 78.55 GB down to 72.48 GB across three configurations. That is a model whose weights dominate.

FramePack's figure moves with its output too, but it is published at a point where a consumer card qualifies, which is a decision about who the release is for.

3What each one leaves out

FramePack states no frame size, so the 6GB figure has no pixel dimensions. Step-Video states no frame rate, so its 204 frame ceiling has no duration.

Neither omission is the same shape as the other, and neither is recoverable from the other columns. Both entries record what is missing rather than filling it in.

  • Hardware
    To generate a 1-minute video at 30fps, 1800 frames, using the 13B model, the minimal required GPU memory is 6GBa floor stated together with the output it buyslllyasviel, FramePack repository / recorded 2026-09-22
  • Hardware
    78.55 GB peak GPU memory at 768px768px204f, 77.64 GB at 544px992px204f and 72.48 GB at 544px992px136fstated per configuration at batch size onestepfun-ai, Step-Video-T2V repository / recorded 2026-09-22

4Sources

Read from github.com/lllyasviel and github.com/stepfun-ai on 2026-09-22. Full entries: FramePack and Step-Video-T2V. What this column is quoting is on hardware. Nearby pairs: Same size, one answer, Floor or builds.