More Shrinkage
That is one of the main ways video codecs shrink file size. Another way is a variety of the "If it ain't broke, don't fix it" philosophy. People notice visually mainly what moves in a visual framework. Once they've taken in the background once, the foreground is everything, where the action is. Only when something moves in the background is it noticed, and not always then, like the gorilla. So, you only have to transmit the background once - then not again until it changes - and only send fresh signal for the changing foreground. That more than halves the size of your signal once again. The viewer gets enough information for the brain to process to keep up a modicum of reality, in a similar way to how the brain works. We don't process every single signal that comes out of our rods and cones fresh every instant, because our brains also have limited processing power. So, if the signal is "good enough" to match our normal brain processing, it appears real.

Fortunately for streaming audio/video, our brains are limited. But they're not nearly as limited as our computer memories or our streaming pipelines. So, another step to shrink down a quickly-disappearing reality must be applied. This goes back to the old Dolby technique of using equipment on the receiving end of a signal to reconstruct it from its shrunken form. Except this time, it will be digital formulae on the receiving end doing the job.

Of Vectors and Layers
This is what Flash animation and also MPEG-4 do. They take a shape in the picture, reduce it to a mathematical formula, and transmit the formula. When the shape changes, the transmission sends out a minimal amount of information about the change ratio and direction, and the change is made on the receiving end, essentially constructing a new picture which hasn't needed to be transmitted again in toto, saving monstrous amounts of space.

Walt Disney did this in an analog fashion in his early feature-length cartoons. He only animated the lead figures, but laid them onto layers of 2-D painted sets, which turned into an approximation of 3D movement when you moved the cameras frame by frame. What Flash and MPEG-4 do is turn each of those layers into initial transmissions with formulae attached and then move the formulae. Flash does it with animated shapes, MPEG-4 does it by segmentizing the elements of the live scene. For a very large, but amazingly well-explained piece on how MPEG-4 and eventually MPEG-7 work, go here.