Video Compression: What takes longer?

Question

So, I've always wondered, when it comes to compression, if it takes less time to encode a video to a smaller resolution, or a larger one.

For the sake of being realistic about the question, let's take an example of a somewhat lossless mov, (maybe mjpeg or prores 422), 29.97 fps, keys set to whatever the compressor wants or 24 if auto is unavailable. I'll do 2 conversions, 1 to 480p @ 800kbps, 1 to 720p @ 1500kbps, both 2-pass, to mp4. I realize that this may be specific to the encoder, but knowing which encoders take longer to do certain things would be good too. If you want specifics, let's assume it's ffmpeg.

If I am converting down to a 480p video at 800kbps, at first that seems like it would make sense that it would be smaller, because it is generating less data.

But then I was thinking, maybe compressing each frame less (if that is in fact what happens), might be faster. So if I were to convert to a 720p at 1500kbps, maybe that would be faster?

I imagine the time difference wouldn't be much between these two specific conversions, but it would be different. What would effect the speed of conversion negatively? The size of the video? The bitrate? keyframes? How would you suggest maximizing the speed of conversion with the least effect on quality?

This is mostly hypothetical, I can't really think of a situation where I wouldn't be able to let a server chug on conversions all night, but I've always wondered if there was something I was doing that unnecessarily slowed down my conversions.

Answer 1

Speculating wildly, you can (naively) account for the time in compression as the time it takes to read the input file, the time it takes to process each sample point from the source, and the time it takes to write out the resulting processed output. If the only thing in that (supremely over-simplified) representation that changes is the output, and the size of that goes down, then your time goes down.

Besides this, the easiest thing to do to solve this problem would be to make an encoding benchmark for yourself, making sure you repeat your tests multiple times to ensure there weren't outside influencing factors.