from 135 megs to 240 meg!?

csonni

Found how to use Final Cut Pro in rotating an .avi Quicktime movie. I imported into FCP and then exported as a Quicktime .avi. Surprisingly, it jumped from the original size to nearly double! The reason for saving in .avi is to retain quality. Exported in the others, but the quality was terrible. What gives?

gordguide

Could be a lot of things, but in short it usually goes something like this:

Compressed formats throw data away. The more compression, the more you may dislike the results.

Uncompressed formats keep all the data. File size will be appropriately large.

Fine, provided you work from uncompressed (original data) to compressed. You stop when your file size or quality compromise is acceptable.

If you are working with data that is already compressed (like an avi) you will have a problem. A compressed avi at 100% DV resolution is an ugly file (no better than it was) at a huge file size. Reduce the file size and you are, in effect, compressing it again.

If you think about it, there is no way to preserve a DV movie's file quality and compress it; you can only pick a compromise and when it comes down to it, the compressed version will look or sound worse, at least at some level.

Even though you started with a compressed file, that rule still applies, with a caveat. Now we have to say there is no way to preserve an AVI's quality and file size if you edit it; you must choose one or the other but cannot get both.

You have two choices when editing an already compressed file:
Keep it at maximum file size, preserving whatever quality was in the original, or
Compress it to a manageable file size, and lose even more quality. No two ways around it.

The moral of the story is use original uncompressed footage whenever possible.

Try this experiment:
Open a JPEG (compressed image).
Do some silly thing to it (eg put a one pixel dot on it).
Save.
Open, do some silly thing again.
Save.

Sooner or later, the image turns to a shadow of it's former self. Each time you save, you re-compress it, and in so doing you throw ever more information away, until at some point there isn't any real information left.

In order to stop the recompression, you need to save at a lossless format. Save the last version as TIFF, and you have an ugly 30 MB file. File size does not equal quality (you can have a perfect, lossless copy of you ugly, multiple-compressed image).

With any compressed format, each time you save, you destroy information (throw it away). Blowing it up won't bring it back, it just makes big ugly files of little ugly files.

Many people wreck their files, over time, by working on the JPEG from a digital camera in an image editing program, and then saving with the same settings (ie save as JPEG). You must save at a lossless format (eg TIFF), even if that means taking a 1 MB JPEG and turning it into a 30 MB TIFF. Quality won't go up, but it won't go down either. Save as a compressed format is the very last thing you do, and it's intended for files that will never be edited again.

Works the same with movies.

__________________
"Being an artist doesn't take much, just everything you got. Which means, of course, that as the process is giving you life, it is also bringing you closer to death. But it's no big deal. They are one and the same and cannot be avoided or denied. So when I totally embrace this process, this life/death, and abandon myself to it, I transcend all this gibberish and hang out with the gods. It seems to me that that is worth the price of admission." —Hubert Selby, Jr.

andreww

Thanks for the insight Gordguide. I don't think that I fully understand why, when he performs an edit on an already compressed file, the file increases in size when saved. Provided he used the exact same settings when exporting you would expect the file size to be close to the original. Does fcp uncompress the file and convert it to DV format and then export as AVI? If so, my guess is that the codec used is simply different than that used to compress the original file.

gordguide

Andreww, think of it this way.

You have a chess board. There are some chess pieces on it; lets say 10 of the 64 squares.

Original Data: We write down a line of code:
<i>Square 1 is empty.
Square 2 is empty.
Square 3 has a pawn.</i>

... etc until we have 64 lines of code. For argument's sake, we will say each line takes up 1 MB of data, so we have a 64 MB file that represents the state of the chess game when we saved it.

We can compress that data file. When we compress it in a lossless format, we do something like this:
<i>Squares 1 and 2 are empty.
Square 3 has a pawn.
Squares 4 has a rook.
...
Squares 49 to 64 are empty.</i>

... etc until we have about 18 lines of code (an arbitrary number, but it represents one for each chess piece [ten lines] and a few more to indicate the empty spaces). At the same file size per line, we now have a 18 MB file that exactly represents the original, and we can re-create the original from it.

We can further reduce the file size, if we agree that the exact details are not important. This would be lossy compression. We might use it if it were important that we knew what it was but not important to know the state of the game. Lets do it this way:

<i>We have a chess board.
There are chess pieces on red spaces 3, 7, 19, and 33.
There are chess pieces on black spaces 4, 6, 8, 10, 44 and 48.
There are 54 spaces that are empty.</i>

Now we have an idea of the original, and if we don't care about recreating the game, it's fine. Using our arbitrary 1 line = 1 MB, it's a 4 MB file.

Original: 64 MB. We can recreate the chess game from it.

Lossless compression: 18 MB. We can still recreate the chess game from it.

Lossy compression: 4 MB. We know someone was playing chess, and we know where to put a fuzzy chess piece, but we no longer know what kind of chess piece it is.

Now, we take the lossy file, 4 MB, and decide to "edit" it. If we save it to preserve maximum information, it goes back to 64 MB but we don't know anything more about the state of the game than the file tells us. So we have a huge file that tells us a little bit about what squares have a fuzzy piece on them. If we try to save it as a 4 MB file again, we are going to have to do something like:

<i>We have a chess board.
10 have chess pieces on them.
We don't know which ones go on which colour squares.
The other 54 are empty.</i>

Back to 4 MB, and we know less than the "original" (compressed) 4 MB file (because lossy compression must always throw some information away; I've arbitrarily thrown away some info for illustration purposes).

Or we could keep the 64 MB file, and know that 4 are on red and 6 on black, which isn't much more information, but is undoubtedly more information than the re-compressed file.

There's the problem; we have a 64 MB file that tells us less than if we had kept the original 64 MB file, but takes up the same space. We should have kept the original and used that for editing. Since we didn't and are working from a compressed file, each time we save we will throw more information away.

So, even though it's a huge file, it's still in our best interest to keep the "dumb" 64 MB file if we intend to work on the file further; if we don't we will just have an even dumber 64 MB file to work with next time we edit.

If you use a program like Final Cut Pro, or PhotoShop, or whatever, it is always going to work with and want to create the 64 MB file, because the purpose of such a program is to preserve quality at all times. You are free to save it any way you want, but if you intend to revisit that file, you either save it in the large size format, or throw information away.

So, there you go. There is no way to preserve all the information, no matter how much or how little there is, if you export in a compressed format. And, if you later use that as the original, there will be even less information after you save it in a compressed format a second time. The key is that all lossy compression always throws away some information every time you save.

It should be obvious that if you do that, sooner or later the file is basically worthless because it contains too little information to mimic the original.

It may be strange to take a 5 MB mp3 and convert it to a 30 MB AIFF file, but the AIFF (or similar uncompressed format like WAV) is the only way to preserve all the information that 5 MB mp3 has. Saving it again as mp3 will throw some of that information away, even though the file size will remain at 5 MB.

File size does not equal amount of information; and lossy formats are not intended for files that need to be revised under any circumstances.

__________________
"Being an artist doesn't take much, just everything you got. Which means, of course, that as the process is giving you life, it is also bringing you closer to death. But it's no big deal. They are one and the same and cannot be avoided or denied. So when I totally embrace this process, this life/death, and abandon myself to it, I transcend all this gibberish and hang out with the gods. It seems to me that that is worth the price of admission." —Hubert Selby, Jr.

gordguide

" ... I don't think that I fully understand why, when he performs an edit on an already compressed file, the file increases in size when saved. ..."

It increased when he opened the file. FCP works with uncompressed data, and it created a file of that size to do so.

Put it this way: take a paper notebook and put an X on one page. Leave the rest blank. Now scan every page. That's how applications work with data; the amount of space dedicated to the blank page is identical to that of the page with the X.

Common sense says scan the page with the X and one blank page, now we have all the data we need. Data for what, though? Not for a murder investigation where the book is evidence; we want to examine every inch, including the "blank" pages, for clues.

So does Final Cut Pro.

__________________
"Being an artist doesn't take much, just everything you got. Which means, of course, that as the process is giving you life, it is also bringing you closer to death. But it's no big deal. They are one and the same and cannot be avoided or denied. So when I totally embrace this process, this life/death, and abandon myself to it, I transcend all this gibberish and hang out with the gods. It seems to me that that is worth the price of admission." —Hubert Selby, Jr.

nxnw

Your compression analogies are not quite right. The concept behind lossy compression is to throw away data that cannot be perceived by human senses. There is a point at which humans are able to perceive the difference, but ideal lossy compression is not distinguishable from the original. Perhaps you might say (in a lossy scenario):

There are white chess pieces on spaces 3, 7, 19, and 33.
There are black chess pieces on spaces 4, 6, 8, 10, 44 and 48.

In a lossless scenario, you would identify the pieces, so you could perfectly recreate the game.

In a lossy scenario, you would throw away the identity of which piece is on which square, on the rationale that you are not close enough to the board to see anything but the colour of the piece anyway, i.e. the reconstructed board would look no different to you whether or not you used lossless or lossy compression.

The problem with repeated recompression is that, while the once compressed copy is not (theoretically) distinguishable to me or you, it IS different from the original. So, if you recompress, you lose additional data and the consequence is that the copy eventually does become distinguishable and inferior. Each time you recompress, you compromise the data and deteriorate the image further.

andreww

Thanks for taking the time to explain that guys. Very informative!

gordguide

" ... The concept behind lossy compression is to throw away data that cannot be perceived by human senses. ..."

The concept behind lossy compression is to throw away as much data as necessary to get to the file size the user demands.

Naturally, there is an attempt to throw away "less important" stuff rather than "more important" stuff. However, the value of the data takes a back seat to the file size required. If you demand that we must compress a 1MB file down to 1 byte, you will get a 1 byte file, regardless of how useful the remaining data is.

Audio codecs do pay lip service to what can and cannot be perceived by listeners. However it is not true that test listeners do not hear the difference between codecs under consideration and the original.

Fraunhofer did not decide on what algorithm eventually became mp3 by using listening tests that compared mp3 to the original, they chose it by comparing it to other lossy codecs under consideration that reduced files to a similar size. They simply picked the least objectionable one.

__________________
"Being an artist doesn't take much, just everything you got. Which means, of course, that as the process is giving you life, it is also bringing you closer to death. But it's no big deal. They are one and the same and cannot be avoided or denied. So when I totally embrace this process, this life/death, and abandon myself to it, I transcend all this gibberish and hang out with the gods. It seems to me that that is worth the price of admission." —Hubert Selby, Jr.

csonni

Very well put- all of you.

nxnw

Well, that's the other side of the coin. Obviously, once you reach a certain point you are compromising quality.

Let's say you start out throwing out data that won't be missed at all, but if you have keep throwing out data to achieve a smaller file size, you throw out what will be missed least.