AI Research Atlas

Movie Gen

Meta · 4 October 2024

30B video model (16 s, 16 fps) and 13B audio model (45 s) for generation, personalisation and editing; announced but never released as weights.

Joint suite covering text-to-video, personalised video from a photo, instruction editing and synchronised sound; Meta reported net-positive human-eval win rates against competing models and published a detailed paper.

Date
Friday, 4 October 2024
Lab
Meta
Kind
model
Access
research preview

Figures

MeasureValueMeasured by
Model sizes30B video, 13B audio
up to 16 s video at 16 fps; 45 s audio
company

Meta said it would work with creators before any release; no public product shipped from the pages checked.

Sources

  1. ai.meta.com/blog/movie-gen-media-foundation-models-generative-ai-video/

This record was checked against its sources on 6 October 2026. How we check

Related