Imagen Video
lastmod 2022-10-07created 2022-10-071 edits
Google's text-to-video AI
Features
- A "Text-conditional Video Generation System" that takes text as input and creates videos with a Video Diffusion Model
- Distinctively, it creates a low-resolution video (24x48 pixels, 16 frames, 3fps) from text, then upscales it by cascading 7 diffusion models
- Final output is 1280x768 at 24fps. Can generate 5.3-second videos
- Paper: Imagen Video : High Definition Video Generation with Diffusion Models