Girdhar et al.: explicit image-first factorization of text-to-video generation

What changed

First generated an image from text, then generated a video conditioned on both that image and the prompt, using adjusted noise schedules and staged training rather than a deep model cascade.

Scientific scope

First generated an image from text, then generated a video conditioned on both that image and the prompt, using adjusted noise schedules and staged training rather than a deep model cascade. Text/image diffusion and image conditioning are inherited. Acquired delta is explicit factorization/training design, not generic video diffusion. Blind human preference results are scoped to tested comparisons and do not establish universal coherence or current rankings.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Text/image diffusion and image conditioning are inherited. Acquired delta is explicit factorization/training design, not generic video diffusion. Blind human preference results are scoped to tested comparisons and do not establish universal coherence or current rankings.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Emu Video: Factorizing Text-to-Video Generation by Explicit Image Conditioning

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction