question about an article from a pro i read

D

deleuze3

Well-known member
Here is a sample of an article i read (my question follow the article):
AUDIO COMPRESSION FOR THE WEB
UPDATE – June, 2010
Recently, I was talking with the tech support folks at Aphex, who make high-end audio gear for network and broadcast work. They told me that MP3 files are optimized for a maximum audio level of -6 dB, while AAC files are optimized for a maximum audio level of 0 dB.
Since much of the audio work that I do gets compressed into both formats, I’ve lowered my audio levels a bit so that they are peaking between -6 and -4.5 dB.
My audio still sounds good and I get nice clean results when compressing to MP3 files.
(For the record, almost all my encoding is the spoken word, as opposed to music, and I use a data rate of 56 kbpsfor MP3 compression of mono audio files.)
NORMALIZING
Based on this, when I normalize clips, I’m now normalizing to -4.5 dB. I also use the same level settings when using the Limiter filter in Soundtrack Pro.


I WAS WONDERING. DOES THE AUTHOR MEAN YOU SHOULD DO THE FINAL EDITING OF THE VIDEO IN MP3 OR DOES HE MEANS WE COULD KEEP THE ORIGINAL WAV FORMAT AND THAT THE WEB COMPRESS THE WAV FILE IN MP3?
FOR BETTER RESULTS, WHAT SHOULD WE DO ?

THANKS IN ADVANCE,
DELEUZ3
 
I think they're referring to the fact the mp3 specification allows for the idea of end user gain adjustment being carried out automatically by encoding into the mp3 details of the mastering level - replay gain was the idea. Not certain, but I think it's built into ipods and similar devices, so relative volume between tracks can be evened out. The idea is to record below the maximum level, and then the player can adjust up or down.

However, unless your production equipment allows you to edit ( mp3gain seems a popular piece of software that enables you to re-encode an mp3 to do different volume adjustment while the file is playing - letting the player turn up a quiet intro, for example.)

There doesn't seem to be any 'rule' that dictates mp3 needs to be recorded at a lower level that I'm aware of. I always master in 24bit/48K or 96K now for audio, and any compressed formats are just produced from the uncompressed version with no changes to level. Doesn't seem to cause any problems for me.

I don't use mp3gain myself - although perhaps it would let the levels on my iPod seem a little closer? I really don't know how widespread this volume adjustment feature is on portable devices, but my research indicates a lot of people are doing it as a matter of course?
 
MP3 is a delivery format and is not really suitable for original recordings or editing. The problem is that data is discarded every time the file is compressed. When you record in MP3, some data is discarded when you make the original recording. Then when you load it into your editing software it's decompressed. But the data that was discarded when the original recording was made is not recovered, it's gone for good. Then you save the edited file back to MP3 and more data is discarded as the file is re-compressed. The data loss is additive with more and more data being discarded each time you go through the cycle. Eventually the data loss becomes noticeable as degradation of the audio quality. It's analogous to editing and saving photos as JPGs - each time you load, edit, and save the file more and more information is permanently lost.
 
Depending on the data compression settings, both MP3, AAC and other lossy data compression formats can clip the waveform in the encoding process. For instance, if the peaks are constantly hitting -0.01dBFS on the source file, they will likely be lots of 'overs' on the MP3.. On 'loud' program material (music CDs for instance) it's necessary to reduce the source file's amplitude to prevent clipping (at least in the encoding process) If in question, I'll open a newly encoded file in an audio editor or metering software to confirm it's not clipped. So I would concur with the article, that maximum peaks of -6 and -4.5 dBFS would likely have no overs after encoding.
 
Steve House said:
MP3 is a delivery format and is not really suitable for original recordings or editing. The problem is that data is discarded every time the file is compressed. When you record in MP3, some data is discarded when you make the original recording. Then when you load it into your editing software it's decompressed. But the data that was discarded when the original recording was made is not recovered, it's gone for good. Then you save the edited file back to MP3 and more data is discarded as the file is re-compressed. The data loss is additive with more and more data being discarded each time you go through the cycle. Eventually the data loss becomes noticeable as degradation of the audio quality. It's analogous to editing and saving photos as JPGs - each time you load, edit, and save the file more and more information is permanently lost.
Click to expand...

Very true - but the word "eventually" is wrong. The data loss is noticeable on the very first MP3 file you make from the original.
 
"The data loss is noticeable on the very first MP3 file you make from the original"
Well.. yes and no. When I was recording, editing and mastering educational seminars for CD (and cassette) I would make archival files on CD or DVROMs, encoding the edited master PCM wave files to MP3, via a WinLAME RC3 encoder at 320kbps, 44.1Khz., CBR mono . In my listening tests, these were identical to the original. Placing both on a timeline and throwing one or the other 180 degrees out of phase, would totally cancel out. Of course, this was spoken word material and did not contain much harmonic or spatial information, AND the MP3 was mono, which would have the equivalent quality of a 740kbps stereo MP3. (if that was possible)
FWIW, the WinLAME RC3 encoder which supported altering scripts which had much more parameter adjustment options than the GUI.
That said, I would also not recommend a data compression format for acquisition or intermediary.
 
Thanks so much for the answers!

I was wondering, i usually shoot in 24bits/48khz.
I heard youtube is compressing any audio in mp3.
Should i compress my audio of 24bits/48khz in mp3 with adobeaudition or i should let youtube compress the file instead ?
thanks in advance
have good day
 
Try it both ways and see which sounds best. I know YouTube used to process every file no mater the format so it was best to give them as full a file as they would take. But I don't know if that is the case now. IF you can give them the format they want and they won't run it again through their compressor then you probably better off doing it yourself. But if there is a chance they will re process you file then give them the best quality you can.

I know that is kind of vauge but their policies change so you need to know what they are doing now. Or just do some tests.
 
According to tests and exhustive experimentation by knowledgeable, uncompromising contributors found on the SCS Vegas Pro forum, the best quality.. after YT or Vimeo does it's (re-encoding) thing, is achieved by uploading an .mp4 (H.264/AAC) file. I concur that the quality is quite impressive, having viewed and tested it on my own material using the recommended parameters.. and software, all of which are freeware aside from a NLE, to edit and render an intial intermediary file. But many alternate NLEs could be used the 'Better' and 'Best' quality methods in the tutorial.
 
Last edited:
You must log in or register to reply here.
Back
Top