Shotcraft
Video generation

FramePack

A next-frame-prediction network structure that compresses input context to a constant length, so generation cost does not grow with video length.

A video generation model by lllyasviel (Lvmin Zhang) · read 2026-08-14 · sourcing low

Up to
60 seconds
Weights
Downloadable
API
No
Commercial use
Permitted subject to conditions

This lets 13B-class video models produce roughly one-minute videos at 30 fps (about 1,800 frames) on as little as 6GB of GPU memory. It is an inference architecture wrapped around existing base weights rather than a from-scratch foundation model.

Commercial use — Permitted subject to conditions

The FramePack repository code is Apache-2.0. Important caveat: FramePack is an architecture and sampling wrapper, and the distributed video weights are derived from a HunyuanVideo-class 13B base.

  • Permission required above a user threshold. Derived weights therefore carry the base model's terms - for HunyuanVideo that is the Tencent Hunyuan Community License with its EU/UK/South Korea exclusion and 100M MAU threshold - not Apache-2.0. The repository README does not spell this out, and this pass did not open the license file on the associated weight repositories, so verify the specific checkpoint you download

The conditions above were extracted from the licence notes recorded during research, and each one shows the sentence it came from so you can check the reading. Read the licence before you rely on any of this. Nothing on this page is legal advice, and this site does not tell you a model is cleared for commercial use.

Specification

Vendor figures, read 2026-08-14Read from lllyasviel (Lvmin Zhang)'s own documentation on 2026-08-14. Vendors change these between model versions without announcing it, so check theirs before you commit a shot to a number. Sourcing for this entry is thin. Treat it as a starting point and verify before relying on any figure. Check the vendor's current documentation

What it takes in
image, text
What it makes
video
Maximum duration
60 seconds
Maximum resolution
not established
Aspect ratios
not established
Controls
reference image, text prompt, target video length, TeaCache toggle, GPU memory preservation slider, anti-drifting sampling, F1 forward-only and P1 variants
Weights
Downloadable
API
No
Pricing
Free, self-hosting only, no first-party API. The headline claim is the point: 'Minimal required GPU memory is 6GB' for generating one-minute videos with 13B models, and 'Laptop GPUs are okay.' Throughput is the tradeoff - roughly 1.5-2.5 seconds per frame on high-end GPUs and 4-8x slower on laptop GPUs.
Released or updated
2025-06

Maximum resolution and supported aspect ratios are not established here. That means no page we could retrieve stated them, not that the model lacks them.

Strengths

  • 6GB VRAM for minute-long video is unmatched
  • Constant-cost generation regardless of length
  • Apache-2.0 code with an approachable one-click installer
  • Anti-drifting sampling keeps long clips coherent
  • Works on laptop GPUs

Weaknesses

  • Licensing of the actual weights is ambiguous and likely inherits Tencent's restrictions rather than Apache-2.0
  • Very slow per frame; a one-minute clip is an overnight job on modest hardware
  • Resolution not documented in the README
  • Quality drifts over long durations despite mitigations
  • Not a foundation model, so it inherits base-model limits

Best for

Where the sources disagree

Recorded rather than resolved. Picking the tidier number would hide the fact that the vendor's own pages do not agree.

What to understand before using it

Explainers naming the mechanisms this model's own documentation mentions.

See it from the vendor

Shotcraft hosts no sample output. Judging a model means watching its own showreel with your own eyes, and a clip re-hosted here would be a clip we chose.

Nearby in the directory

Adobe Firefly Video Model

The Adobe Firefly Video Model generates video from a text prompt, optionally guided by an image used as the first or final frame. Adobe's Firefly Serv

Video generation

Allegro

A small text-to-video model comprising a 2.8B VideoDiT and a 175M VideoVAE, released under Apache-2.0 in October 2024. It generates 88-frame clips at

Video generation

Amazon Nova Reel 1.1

Amazon Nova Reel generates video from text prompts or from text plus an image, and is available only through Amazon Bedrock. AWS documents output at 1

Video generation

Sources

Read on 2026-08-14. A source that has changed since then will not be reflected here until the next pass.