Linux

Since the end goal is to post a video to YouTube, you will have to create a video file. Personally I would probably just be lazy and upload the large file, since YouTube is going to reconvert the video anyway.

That said, to optimize the file you need to know how videos work, specifically key frames. Speaking generally, when a video gets encoded, it doesn't add the whole image for each frame. Instead, it only does that when the current frame is a key frame, and then only stores the difference to the previous frame for every regular frame. There's a lot of different strategies when placing keyframes, like every X seconds, when the scene changes, or both. This is usually you can change somewhere in the encoding settings of the application you're using. You will need to use a codec/format that supports interframe compression though, so avoid AVI and MJPEG.

So the TL;DR is: Try to decrease the amount of key frames as much as possible, maybe even down to only one if possible.