Medium Image & Video Gen · 1 min read
Imagen Video and Phenaki: Google answers on text-to-video
In one sentence Six days after Meta's Make-A-Video, Google publishes two text-to-video papers: Imagen Video, a cascade of diffusion models up to 1280×768 at 24 fps, and Phenaki, which generates variable-length video from a sequence of prompts.
Reading level
Six days after Meta, Google shows two answers on text-to-video, published on the same day.
Imagen Video extends the approach of its image model: it produces clips of about 5 seconds in high definition. Phenaki aims at length instead: it follows a sequence of sentences changing over time and can tell a short story, with lower visual quality.
Both remain research: Google does not release the Imagen Video model or its code.
Companies
Google Research
Tools
Imagen Video, Phenaki
Tags
GoogleImagen VideoPhenakiText-to-VideoDiffusion
Sources