h.265 - Do large video streaming services use intra-video compression? Are there any tools that do this?

Question

So video codecs like h.265 use intra-frame coding to compress a video by leveraging the information that from one frame to the next, only some pixels might change or a group of pixels may stay the same color but change position, e.g. an object moving from left to right.

But how about intra-video compression? Large services like youtube or even GIF-based websites must have hundreds or thousands of versions of popular content, e.g. cutscenes of video-games in Let's Play's or reviews, animated memes which may only differ in the caption.

At work we might also have a scenario where we capture tons of video data for a dataset, with only minor differences between individual videos in e.g. the background or the color of an object.

One idea could be to split all videos into frames using e.g. ffmpeg and then somehow modify h.265 to not only compare previous and later frames and their differences, but also the frames in other videos, and the resulting format would basically only reference a database of key-frames and differences.

Has any of this been done before? Would this even yield smaller files as compared to regular h.265 compression?

question from:https://stackoverflow.com/questions/65893996/do-large-video-streaming-services-use-intra-video-compression-are-there-any-too

Answer 1

Waitting for answers