The video's first frame is a key frame (i.e. it contains the full picture) and with most codecs the following frames contain
only the changes to the current picture on screen.
When you open the video file in a player (e.g.
MPCH, MPlayer, VLC, ...), it will play back the file without problems. You can seek forward (jump to a later point in the timeline than the current position) and the player will try to apply all the picture changes up to that selected point as fast as possible.
But if you want to seek backwards (jump to a previous point in the timeline), the player has to start at frame 0 again and apply all the changes from there. So jumping from 1h00min30sec to 1h01min00sec is no problem, but jumping to 1h00min00sec will take a long time because one hour of video data has to be played back to reach that point.
That's why normal video files need keyframes (full pictures) in regular intervals. If you use the video dump only for transcoding or for uploading to YT then you don't need keyframes.
Btw. I forgot something when I made the above post: Note that some codecs
always store keyframes for every frame (which is one reason why they produce so large files). I think Camstudio is one of them, ZMBV isn't, and Lagarith has the ability to use "null frames" when a frame has no changes from the previous one. So in case of Camstudio that option wouldn't make a difference.
Also, the method you told me about to try on Handbrake keeps giving me this error:
