Movie Gen Paper Access and Rights Record

Movie Gen Paper Access and Rights Record

Research claim

Meta's Movie Gen publication presents a family of media foundation models for text-to-video generation, personalized video, instruction-based editing, video-to-audio generation, and text-to-audio generation.

The largest reported video model has 30 billion parameters. The paper page describes generation at 1080p, up to 16 seconds at 16 frames per second, using a maximum context of 73,000 video tokens. Meta's October 4 announcement describes a separate 13 billion parameter audio model and audio generation up to 45 seconds.

Access boundary

The 2024 research page provided the paper, selected examples, and a creator feedback program. The Blumhouse announcement documented limited creator access. None of those records established unrestricted public access to original Movie Gen code or weights.

Meta launched preset-prompt video editing in 2025 and described the feature as inspired by Movie Gen. That wording matters. It supports a research-to-product connection without proving that the original research models were released as a public product.

Rights and production boundary

First-party examples and evaluations show what the research team selected and measured. They do not resolve rights to an input image, a person's likeness, training material, generated output, music, voice, or downstream commercial use.

A production review still needs current product terms, geographic availability, provenance controls, consent, security, retention, output review, accessibility, latency, consistency, and an exit path. Movie Gen's paper is research evidence, not a blanket rights clearance or production approval.