audio mixing and normalizing

fp615 wrote on 11/15/2013, 6:12 AM
For the show I asked about 5.1 audio production, I have a wav recorded by a zoom with mics in the stage ceiling.
I did first quick'n dirty editing. The comments are that it is ok except for the audio: it is too "clean". Used to sound from cameras set in the middle of the public with childer crying, people commenting or moving, etc this time the audio is "too clean"...
You can hear public clapping hands or laughing very low.

So my idea was to mix the audio track from one of the cameras to give audio more "presence". (putting "public" track in the back channel of 5.1 is an idea, but playing on standard dvd do I have a mix ?)

A friend of mine told me that I should normalize audio tracks first at -3db and use "actor" track as reference. Then use the "ambient" track upping levels when needed. The 2 tracks, summing, won't exceeed 0db....
I don't know if it is a correct method. If it is not wrong, can you please tell me how to proceed ?



musicvid10 wrote on 11/15/2013, 8:54 AM
It's an OK approach, however your targets are a little off.
Average RMS levels would run -12 to -6dB (which is pretty loud), these are entirely different than peak levels which is what you see on Vegas meters. Peak levels are pretty useless for setting loudness levels.

The RMS Normalization settings in Sound Forge are a good way for beginners to get tracks in the right ballpark without causing a lot of stress. Start at about -9dB with peak compression turned on.

As far as "too clean," mixing in ambience can help, so can a tiny bit of compression and reverb in your surround project. Run the voices front dead center, music, etc. to the sides, and ambiance to the rear, down -12dB from the fronts.

Remember these are only starting points, and you will want to render lots of samples to see how it sounds on your home Dolby Surround system. A good way to compare is use a movie that you like as a reference point. Any average levels approaching 0dBFS would be 'way to hot.
fp615 wrote on 11/15/2013, 10:59 AM
Thanks for reply.

Can you add a bit of information on how to normalize at -9db ? I don't know how to explain... let's try.

This was a show of children, from 6 to 13 year old. Some action was performed on stage, some down the stage (and audio level is lower). On stage, some have a loud voice, some have low voice...

I have Sound Forge 10 studio. I think that normalizing is done with: open the wav, process -> normalize (not RMS Normalization), set the slider, press OK.

I opened the wav. Inside there is everything. From the little children speaking at -18db to the actors screaming clipping audio...

So I did 2 tests:
a) select the whole file, normalize at -3db (just for a test)... peaks were moved to -3db but all the audio file was compressed... low voice actors have lower voices now...
b) select small segments.... imagine actor+claps+actor+claps = 4 segments. Select one and normalize, select the other, etc. Unfortunately, when you play, you hear when you switch segment... mics have "clean" signal but you may hear some public presence anyway, and it is amplified too... I hear a pressure in the earhones... I don't know how to describe...

So, probably I don't need a simple normalization but a more complicated audio recovery.

So I loaded the project in Vegas and found a camera audio track good enough for music and ambience effects. I tried to set a high gain to the actor track and insert ambience effects when needed. It seems to work ok. Not perfect, but ok.

willqen wrote on 11/15/2013, 11:41 AM
You had it right in Sound Forge.

Process - Normalization.

Just switch it from peak to average RMS levels, make sure that "if clipping occurs" is set to "Dynamic Compression".
I would set my attack times fairly short, say to between 6ms and 10ms because speech is involved. You may choose differently, perhaps even shorter times. You will have to experiment.

Release times will also have to be short, but not as short. Say 40ms to 70ms. Any longer and you will begin to experience audio artifacts. Again listening to the result is the key.

If you set your peak value to -9db and switch to RMS as I've outlined above, and set your attack & release times properly, you should have a fairly good audio track that is compressed and no section should be louder or softer than another.

Oh yes, be sure and do the whole track. not just sections ...

Good Luck, let us know how it turns out ...

fp615 wrote on 11/19/2013, 4:18 AM
> You had it right in Sound Forge. Process - Normalization.
> Just switch it from peak to average RMS levels

I have "studio" version... I don't find this "switch" and the other settings you talk about.

> Good Luck, let us know how it turns out ...

Non really well... :-)

farss wrote on 11/19/2013, 5:01 AM
I'm currently in the process of editing one such project and I do a lot of them. Musicvid is on the money however TBH I've never used normalisation.

Certainly a feed from a desk will sound "flat" because a room adds reverb and the feeds I get are before the outboard compressors and equalisers. Adding a little compression and THEN a little reverb will get you what you want. I use ambient sound from the cameras or the other channels on my recorder and a room mic with great care because of possible phase problems. If I must use them I exclude the lower frequencies from them.

Just remember with audio a little goes a long way. No matter what FX you're adding once you can hear it's impact it's probably the point where you should wind it back a notch. Everything downstream will add more compression.