How to prevent Speech Volume Leveling in Adobe Audition from amplifying noise?

Question

I am recording some speech (for voiceover purposes) and have noticed two issues with the input that I wish to correct:

1. There is a persistent hiss. The Noise Reduction effect appears to deal with it, more or less.
2. Since I cut and paste the final output, my audio volume is not always constant. I am trying to use the Speech Volume Leveler effect to correct this.

The result that I get after applying these two steps is that moments of silence in the audio file get their light remaining hiss greatly amplified. My desired goal is to leave silent moments as-is and just level the parts with meaningful content.

Adobe documentation appears to support the idea that background noise is not meant to get louder, so this is perpelxing.

The Amplitude and Compression > Speech Volume Leveler is a compression effect that optimizes dialogue, evening out levels and removing background noise.

Here are some example files that demonstrate the issue.

• Raw input
• After Noise Removal
• After NR + Speech Volume Leveler

Am I doing something wrong process-wise?

Answer 1

First of all, you are making a fundamental error which is resulting in a higher noise level. This error is with your recording levels. The level of your voice is too low during recording.

Secondly, you have DC bias(offset) occurring somewhere in your system. This lowers the highest level your voice can be recorded at. Again, the result is a higher noise level.

If you sort these two things out, you should have a much better starting point to work from. From there, I suggest you use the 'Noise Reduction Process...' with a 'noise print' of an isolated section of noise from your recording.

As for the SVL, read my answer to this question regarding scrapping it.

Good Luck.

Answer 2

A "speech volume leveler" is just a fancy name for some proprietary algorithm they've cooked up to try to make things simpler for the amateur user.

The basic, building block tools to control your levels here are a compressor and an expander. You'll probably need both. Apply the expander first. If you set the threshold and other attributes correctly, all signals that are louder than the noise floor will be boosted, and in comparison the noise will be much less present especially in quiet moments between sentences. This will still leave some parts of the audio louder and others (your initial issue, that the level is not always constant). However a simple compressor can help to address this. Adjust the threshold till you get your louder parts a bit squashed, I suppose. I use volume automation and apply compression as lightly as I can get away with, as it will affect the character of the sound.

Note I did not download your original files because I didn't desire to install an unknown downloading client on my computer. My advice though is basic common sense audio practice that can come in handy in dealing with any vocal recording with any level of noise.

A noise reducer is another good plugin to apply lightly but it is a filter, and it won't in itself solve all your dynamic problems.