Definition
Plain language
A translator that squeezes each video frame into a compact summary and later paints the picture back out.
As stated in the literature
An encoder-decoder pair that compresses frames into a low-dimensional latent representation and reconstructs pixels from it; here built on frozen DINOv3 features with a learned bottleneck.
Also called: codecs
Why it matters: Compressing frames into a compact form lets a system store and generate video far more cheaply than working with raw pixels.
For example, instead of storing every pixel of a video frame, a codec keeps a small summary and later rebuilds the full picture from it.
Heard on the show
“The codec, a translator that squeezes each frame into a compact summary and paints pixels back out of it.”Episode 206 — How Four-Second Clips Become Hours of Playable AI Soccer