Can't get audio to normalize for my life...

Sticky Fingaz wrote on 2/29/2004, 7:04 PM
As some of you may remember I am making a music DVD for a group by the name of ONYX. Naturally, all 4 of their albums that I am ripping CDs from to splice into the videos have different levels of audio, and like you all rightfully assumed, the newer albums are louder.

I have tried everything I can humanly think of to correct this. I tried LOOKING at the audio and making them appear similar in volume with a WAV editing program. I've tried right clicking the audio in vegas and choosing normalize. I've tried BY EAR adjusting the DB volume in Vegas and I still end up being wrong no matter how hard I try. And today I converted all the rips to MP3's so I can use MP3GAIN to normalize all of them, and look at the difference:

http://members.aol.com/silathis2/why.mp3

I mean listen to the first clip and the second. Does that sound like the same volume as anyone else? I think not.

It is driving me insane. My number one goal with this DVD is to make sure no one has to get up and fiddle with their DVD player everytime a song changes. I really need some help from you guys...it is driving me mad.

Comments

farss wrote on 2/29/2004, 7:14 PM
I suspect the problem is that the later material is much more compressed and hence sounds louder as it has more energy. The scientific way to correct this is to normalise and then measure the RMS value of each clip. Say your quietest one comes in at -30dB and the loudest at -15dB.

Then you need to turn th loudest one down 15dB to get them to sound the same.

Even this is not 100% surefire. As you change the overall volumes of sound the relative levels between two clips can seem to sound different. This is because the ears response changes as the SPL increases. Also there's psychological factors at play, if it's music you hate it always sounds too loud!

Best you can hope for is getting the RMS values to be around the same. If you don't have SF to make the measurements then the 'fatness' of the wave on the TL is not a bad estimate.
Sticky Fingaz wrote on 2/29/2004, 7:19 PM
Yeah I used the "Fatness" method, worked probably the best, but still not good enough....
TheHappyFriar wrote on 2/29/2004, 7:55 PM
Yeah, for what's "wrong" with those songs you'll just have to mess with it. The normalize is doing the right thing: taking the loudest part of the song and making that 0db (and increasing everything else the same amount).

You couldtry doing an EQ to the "quiter" one. It sounds like the first clip has more bass then the 2nd. Increasing certain frequencies might do what you want.
Sticky Fingaz wrote on 2/29/2004, 8:28 PM
It appears I may have to do this by ear.

What I need to know now is, I did this process:

Ripped their albums, used a WAV program to make them all LOOK the same volume. Opened them in Vegas, tweaked with the DB a little more, saved as WAVs, and will open them in Vegas one last time when the video gets added to it. Is all this DB messing around going to alter the quality at all? eg: Hurt bass *punch* or sound quality?
farss wrote on 2/29/2004, 9:36 PM
If you're only changing level then you're not changing the 'sound' of the clips. However taking level up and down and rendering out does slowly intoduce more artifacts. Keeping things in 3dB increments will minimise this as bits aren't being dithered. Also as Vegas does the calcs in floating point numbers the risks are small.

However it's still best to just do it once when you know how much gain you need.
Playing around with Eq or compression is another matter, now you're remixing the track and from what you've said this isn't hat you should be doing.

I thinki you've simply got to accpet the fact that you'll never get say a piece by Brahms to sound like it's at the the same volume as a bit of techno.

Also Normalise has NOTHING to do with loudness! Its sets the peaks to -0.5dB. Two pieces of program can both have their peaks at that level and yet have a 30dB difference in loudness.