Where does the 'extra' data come from when transcoding video?

Question

This is something I've never quite understood, but have always wondered. Forgive my basic knowledge here, but I've tried to research this without a specific answer as to what is going on.

When I convert from an .MTS file to ProRes422 HQ, the file get's huge. Why is that? Where is that extra data coming from if it's not there to begin with? i.e. higher bitrate. I understand the codec has a standard bitrate it will always maintain, but why do it if the original file bitrate is so much lower?

For example, a stock GH2 mts for me streams at about 22 Mbits/s, I convert it to ProRes422 HQ, that same file is now playing at 216 Mbits/s. - With no discernible image quality difference.

What is going on in the codec that makes the file so huge? Other than workflow issues, why would you want to transcode from MTS? Does converting it to ProRes (or another codec) 'Unpack' the data in a more efficient way?

Why not use something like ProRes422 LT if you're starting from a much lower data rate like a native MTS file provides?

Answer 1

When video or audio data is compressed, it works by two different mechanisms. The first (which is the only one used by lossless compression) is to look for patterns that are repeated. When patterns are found, it can store the data once and reference the stored copy.

The second is only used lossy compression and involves discarding the least significant parts of the information. This can be done in several creative ways such as being more forgiving in what pattern can be matched, but the loss of information generally ends up equating to a slight loss of detail.

You can think of it kind of like blurring a photo. You still have just as many pixels and just as much "raw" data, but the amount of detail stored in that data is reduced, and thus it can be compressed further because it fits patterns better. When you move from the high compression, lossy format to a lossless format, a large number of those savings are no longer applied, so it stores the raw "but fudged" data again under a new system that isn't capable of as much compression.

Basically the video is uncompressed from your original format to form a file of 1920 by 1080 pixels (assuming it's a 1080p file anyway) and then it is recompressed differently to become a ProRes file. The file size is large because the ways that it will try to simplify the file are simpler, but the information that is recorded is still the blurred copy that resulted from decompressing the smaller file.

This is also why it is important to work with the highest possible quality files until final output because repeated recompression of files will result in additional quality loss every time.

Answer 2

Some filetypes are compressed, so when you transcode, your software will first decompress, so the interim file is going to be much larger. If you then convert to a filetype which isn't compressed, then you stay at this large size.

It won't have any better quality, as you can never improve quality by transcoding, and in fact it is likely to worse - possibly not by much, but sometimes transcoding really alters the quality.

A key difference, and the reason why uncompressed filetypes are used is that they can be played back on devices with low compute power. If you have used strong compression, the device has to do a lot of work to decompress fast enough to play it back - so you may get lag, stuttering, glitches etc.