Syntrillium Backup Forums :: View topic - Need expert advice on complex sweetening problem!

PhoneGuy · Posted - Thu Apr 24, 2003 5:35 pm

I've got old taped phone interviews that need to be sweetened. One voice is very loud, the other is very soft. Is it best to amplify or normalize the soft, or to reduce the loud and then amplify or normalize? Or, is there another way to do this...hard limit??? And...should the noise and/or hiss be removed first or after the amplifying or normalizing?

Many thanks for your advice!

jonrose · Location: USA

Hi PhoneGuy,

Once again, this is very program dependent, and you're on the right track - obviously you understand what needs to be done. I think the best thing to do would be to experiment a little to see what works best for the situation. As far as pulling out noise pre- or post-amplification, this sometimes depends a lot on the source, too. Can you post any short samples somewhere, or is the material confidential?

Best... -Jon

PhoneGuy · Posted - Thu Apr 24, 2003 6:22 pm

protovack · Location: USA

I would hard limit..because it would be the fastest way. If you wanted to take the time to individually amplify the quiet parts, you certainly could, but you might end up with small volume swells all over the place unless you are VERY accurate with your selections. Try a moderate hard limit and see what happens.

Good luck

rpc9943 · Location: USA

uh am i crazy here but wouldnt it be best to use a compressor?
then noise redux

ron cavagnaro

jonrose · Location: USA

Well, if time is of the essence, I suppose one could try hard limiting... but this seems a bit extreme, and I would probably try to take out whatever noise I could before doing so. The hard limiter can be a very unforgiving compressor (whereas, the ratio is infinity) and it can squish anything over the threshold pretty hard. If the softer and louder voices are at very disparate levels, then using a gain setting in the hard limiter that amplifies the softer passages adequately might really trash the louder voice, as an example. But it's really hard to say without an example.

I still think you will have to experiment, but I suppose you could try a bit of noise removal first, however you tackle it. This might save you some grief later, if using (any form of) compression.

Good luck! -Jon

Noddy · Location: Australia

Phoneguy,
Here's a totally different suggestion... why don't you throw the wav file into the multitrack twice.
Cut one version of the wav to just leave person A.
Cut the second (adjacent) wav to just leave person B.
Boost whoever is the soft speaker up to the level of the loud speaker using track volume.
Then, go and smooth out the edit points with some block automation.
Mix it out to a single file, and THEN run a mild compressor over THAT, to pull it all into line.
That is of course, IF you have the time!
Cheers,
Bruce.

Bobbsy · Location: United Kingdom

Okay...going by instinct a bit here...

The problem with compression on inherently noisy files (like phone calls) is that it is very easy to get "pumping" on the background noise. Therefore, I'd be tempted to run as much noise reduction as you can get away with as a first step. At that stage I'd move to compression and simply experiment to get the best balance between evening out levels vs. jarring changes in the background noise.

Hope this helps.

Bob

Craig Jackman · Location: Canada

You are dealing with a telephone recording so it's never going to be perfectly clear. Run noise reduction if you want, but it won't make that much difference. Run some eq (FFT filter) to eleiminate whatever isn't there to begin with. CEP has some telephone presets that will help clean things up nicely. Boost the mids some more to increase inteligibilty. I would create some envelopes, and apply them over the quiet bits to swell the volume to bring the 2 sides closer in amplitude. Then compress at about 3:1. If needed you can Hard Limit after that, but set the boost amount much lower than standard ... 2 or 3db depending on where your levels are at.

It's a telephone conversation. Whoever was at the recording end is going to be recorded way louder than the other side. There will always be a difference in tone and level.

VoodooRadio · Location: USA

FWIW, I (personally) would normalize the entire file, then get a section of silence and run the "get profile" function in the noise reduction and get as much noise as possible (without creating any artifacts in the voices) and then manually finish getting the "crap" out of the file. Now, with everything "clean", you could do any of the above suggestions. Again, (personally), I would probably use the amplify function and turn down the loudest voice to an exceptable level. At that point, you could weigh whether or not the file needs any further processing. As JonRose alludes to in his post.... "you will have to experiment". The best thing about it is.... having the "UNDO" feature enabled. If you try something and don't like it, undo and start over! Wink

Kensho0mu · Posted - Fri Apr 25, 2003 8:41 am

I have a preset I've created that you may find useful. Select the wave form in question and go to Transform\Amplitude\Dynamics Processing. In the Graphic tab, left click on the yellow line to creat a node and then right click the node to input settings for that node. The preset has three nodes. The settings are: Input -50 Output -50, Input -20 Output -5, and Input 0 Output -5. This results in expansion and compression with the following ratios:

cmp infinite above -20dB
exp 1.5 : 1 below -20dB
flat 1 : 1 below -50dB

You may have to tweak it some for your particular files. I hope it helps. Good luck!