Lessening background noise when Amplifying a sound file (post noise reduction)

Question

I have a video I have been tryng to clean up, that was filmed in an environment with some amount of background noise.
I managed to get the audio cleaned up to a leveI was happy with (but still retained a minimal amount of almost unoticable noise to keep the voise still sounding ok).
It sounds fine and perfectly loud in headphones but on some computer speakers even turned up to full volume you can barely hear it. (in the future I will try to film with a higher volume)
I attempted a few different methods to Amplify the sound slightly but this also increases the reminant of background noise.
     Along with the basic Amplify, and Normalise I have also tried using a Hard Limiter (I don't know if I was doing it right though as it also increased the Background noise)
Can someone provide any pointers to only amplify the Voice and not the background noise? or possibly even suggest a way I can clean up the reminant background noise without causing the voice to become hollow and Tinny? Obviously redoing the Noise removal process just makes the voice more Hollow.
In the future should I Amplify first and then run the noise removal?
I am using Audition CS6

Answer 1

In Audition, Noise Reduction always seems to work best if applied in small steps, with different settings for each. The setting that makes absolutely the most difference is the FFT size you use. The thing to do at the outset is to use the default setting (4096 I think) but don't take too much off with that. Then change the FFT size to the maximum setting, take a new noise print from the results of your previous pass, and take the rest off with this setting. By doing this, you'll find that you can take away more noise before artefacts start to creep in.
It doesn't actually matter about the level of your signal when you apply NR, but to be sensible about it, in order to be able to monitor the results effectively, you should normalize it first, to around -1 or -2dB. The concept of amplifying a signal without amplifying the noise in it implies making the dynamics non-linear, and whilst that's possible to a degree with the Dynamics Processor, it's generally not a good idea - it makes everything seem completely unnatural-sounding.