Mixing sound design with poorly recorded room audio/interviews

Question

Apologies if this has been asked before. I'm designing around voice tracks, some of them recorded in studio and some with a portable recorder on location (talent's house). I find it hard to strike a pleasing balance between the location recordings and my designed sound.

For starters, the drop-ins/outs are so obvious. But besides that, my main concern is that there's already so much background and room noise in the recordings that adding sound design just muddies the mix — and then, if I add music or sound beneath the voices, I need to compress them. That of course brings up the background noise even more.

I've used X-Noise to kill a bit of the ambient noise, but as you'll well know, it doesn't sound very natural and too much of it just makes everything sound like a 28.8kbps Skype chat.

Do you have any tips? If I could turn back time, I'd ensure everything got recorded in studio save for small segments of location. This, by the way, is for a radio/podcast documentary.

Answer 1

I think you'll need a balance of effective noise reduction (as everyone else mentions), as well as fx/music that's been selected to cut through the noise floor (because of its frequency characteristics and transient content). Your fades between music and voice will be important, as you'll have to avoid your noise floor noticeably "pumping".

One trick you might try is to use eq to cut 1.5-3kHz, and 200-400Hz (adjust for taste) in any music that goes over the voice. This lets you pump the music more, while still having the voice audible.

Most b/g noise is pretty broadband. You can afford to cut frequencies below 80-100Hz, and above 17kHz. You can also put a 3(ish)dB dip around 800Hz-1kHz without it affecting the voice noticeably. I'm making an assumption here, but i'm guessing you'll have trouble making the more subtle FX poke through, and stuff like synth pads could get buried.

Hope that helps!

Answer 2

Basically you have two options here. Either remove the background noise with top notch restoration plugin such as Izotope RX or add noise from the poor quality recordings to mask the fades. I think in this case the best results would come from proper audio restoration and (noise reduction, declick, spectral repair) and noise gate combo, but sometimes you have to make ends meet and do what ever you have to with what ever plugins you might got.

Answer 3

Hi there! While I prefer RX, I have used X-Noise on several occasions and, when applied correctly, it is one of the more natural sounding NR applications out there. Also, if you have Z-noise in your Waves bundle, this works in a similar way to the RX spectral repair and will be a great help with impulse noises.

The keys to successful NR are: 1) Work in small steps and do not try to remove all the noise in one pass. 2) Don't try to remove too much! Always AB with the original and the moment it starts sounding artificial you've gone too far.

Answer 4

the order for noise reduction is: filters (i.e EQ), interpolation (de-clicking, etc.), and broad-band noise reduction. you want to do as much as you can with the first two before applying noise-reduction processes like X-Noise, that way that particular algorithm doesn't have to work quite as hard. that means it won't color your audio as much. granted, you'll have coloration from the filters and interpolation (if you even need that second one), but it will be less offensive than that introduced by "noise reduction" processes.

also remember that you can usually achieve slightly cleaner sounding results by using multiple passes/instantiations at mild to moderate settings than you can when using heavier settings.

you can probably get some additional help by using multi-band expansion. if you've got Waves C4, set the range on the band your working on to a positive number. that band will become an expander, instead of a compressor. if you don't have Waves C4, you can "build it yourself" by setting up multiple auxes...each with a band pass EQ (frequency ranges as you desire) followed by an expander. either method will let you tailor the behavior of specific frequency ranges, and can help you mitigate noise.