Slashdot Mirror
Normalizing Music?
Beans asks: "I have a couple classical music CD's which I listen to at work, and use for putting the baby to sleep. I can never find the correct volume, I can't hear soft spots, so I turn it up, only to have a rising crescendo rouse the baby, or at work, have co-workers glace over. What is a good way to normalize them (read on for what I mean by normalize)? All of the normalizing software I have seen uses the entire song for the window of normalzing. Basically makes determines a static value required to get the average volume of the song to the user defined level, then applies that value to the entire song. What I need is something that normalizes over a sliding window, or say 5 seconds, or whatever. In effect making soft spots louder, and crescendo's quieter. Not the way the music was intended to be heard, but perfect for music-at-work, or putting kids to sleep. Does anyone know of any software that does this? On a side note, I work for a Seismic processing company, and we do stuff like this all the time on Seismic waves, not sound waves. If I can't find any canned software to do this, I may modify some of our code to work with WAV files, but I don't want to reinvent the wheel."
That's not normalizing, that's dynamics compression.
Normalizing scales the entire file so that the highest sample is the highest value that can be represented. For many sound formats, this is 32768, for 15 bits. (1 bit left for the sign). This lets the sample use the full range. However, the volume of the file is based on the value of one sample.
What you want is to amplify different parts of the file so that the level of the song is more consistent. For example, sample every 1/4 second and adjust the level based on the average of that sample. A typical value might be 2:1 compression above -10 db. In this case, if the volume is above -10 db, the amount above -10 db is cut in half. -10 stays at -10, but -6 is scaled to -8, 0 to -5. (0 db is the highest value that can be represented.) There is a lot more to it, including look ahead, recovery time, multiple levels, etc.
It is similar to riding the voume control manually and turning it down for the loud parts and back up for the soft parts, but very quickly.
After dynamics adjustment, you usually normalize the sample.
I edit speech audio for my church and deal with this all the time. I use Cool Edit 2000 and it is great at this. However, it is no longer available. But most audio editing programs can do this. Try Audacity.
Headphones.
True - you don't see scoring for a cannon even in the hardest Metal... but sometimes you want to be able to hear it over road noise and/or not scare the baby.
You could also argue that we should all set our eq's flat, or at least only for equipment compensation because that's "how the music was intended." Me, I like to boost the mid-bass a bit and don't apologize for it.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
CoolEdit was bought by Adobe and is now available under the name Adobe Audition. Perhaps you were referring to the lack of a budget version of the software, however, in which case, yeah, that kinda sucks.
Replay Gain is designed to level all the tracks from a CD (and infact many different CDs if you do it to all of them) so that they all have a mean volume of about the same, but it works on the track level, not a 5-second floating window level.
I acknowledge that not everybody (including me) is an audiophile, but the recording engineer went to some effort to make a recording that, when played, would reproduce the original performance as closely as possible.
If you ride the volume control (or use automatic gain control to do it for you) or use dynamic range compression (a different animal from digital compression), you're compromising the music.
I'd no sooner do this than bleach my favourite painting.