Noise Reduction Workflow

[a.finley]

I'm working with dialogue and wondering at what point are noise reduction plugins (e.g. xnoise, rx denoiser) usually used?

For example, are people applying this destructively at the clip level before making edits, or in the effects bin on the track being edited? Something else?

Is there a difference between "crossfading two denoised clips" and "denoising two crossfaded clips" ?

How do people accomplish 2-pass noise reduction? At some point you have to be destructive right?

[eoats]

My method is to 1st up have a copy of all my 'clips' (either the AAf or OMF tracks) inactive & hidden in case I want to get another copy or check sync.

Then going through choosing the mic/channel & trimming, scraping (aka fade in/out). A rough cut, if you will, of the pieces to listen what's there & what's missing.

Next pass through is where I do the De-Noise, De-Clip, et al... when I have maybe found some air to fill with and found some other words/pieces to help smooth things. If possible I'll use an Audio-Suite or destructive process to cleanse pieces. When there's a lot of complex edits for "ah's" or clicks then I'll make a copy to fall back to & move that onto an 'X-Track'. Usually I'll have the 'air' handles as part of the piece being processed, unless it's going to need some EQ or other work to fill in.

[Jfriah]

When do I usually apply noise-reduction? After my breakfast.
(badum-pssh!)

+1 to checking sync after hard processing.

If I'm feeling really good about the edit at the time of NR, I'll do the whole 'piece' as one continuous file. BUT---you lose your handles when you hard process, so depending on your style of editing, I find it a real PITA to go drag a bit on either side, just to give myself 'room to finesse' in and out of the line before NR processing.

Good luck.
Jeff

[a.finley]

Thanks for the tips! Really helped improve my workflow on my last project.

I hadn't even thought of checking sync after NR

[mikevarela]

Like up above,

create muted/un-voiced duplicates of the material your gonna work on - for safety.

I usually attempt to do a rough balance mix of the Dx tracks, then go in by scene or by worst offender and find the right mix of noise reduction to original - then save the setting with the name of the scene and movie so you can apply consistent NR to other clips. I usually pull out the handles a bit and then audiosuite (destructive), however, I usually go for a C4 (multiband expansion) first, as I find it generally works better than the x-noise stuff I have - no digititis. Although you can work it too hard and lose top end.

Usually you have to make a call at how far to process, often times it's ok to have more original than you thought, especially when there's some ambiance or BG material to fill.

Multi-pass usually means working the material two or three times, but less hard each time.

I'd also suggest trying everything EXCEPT NR plugs first... i.e. EQ, expansion, leveling, alts etc. Then attack with NR if you need too.. also communicate with the director about ADR as well, nothing sucks as much as working your tail off then finding they're gonna loop the material

[ErikG]

BUT IF you are going to use NR processing, most of the time you are better off doing that BEFORE applying any other kind of gain control (even destructive clip gain in PT).
Any change in gain be it via a gain plugin or C4 or EQ if that is not consistent through the scene, then the noise floor will vary too much and the process becomes much more audible.

Doing the NR before applying any other destructive processing is my suggestion.

BUT NR processing can easily hurt you original recording, and its sometimes a little too easy to remove to much so the end result sounds lifeless or even extremely digital.

[Bill@AudioVision]

Here's why I love the Fairlight for this sort of thing. I don't worry about making backups of the original material, it's done automatically where needed.
I have 4 bands of clip based parametric EQ and gain that is real time non-destructive... so I can go back to any of it to tweak.
If I use a NR plug-in, I always do it clip by clip. The Fairlight keeps the original clip available at all times in the layer underneath. I can instantly A/B the two at any time later to make sure I'm happy. Any EQ or gain processing caries to the new rendered clip on top, and can be tweaked as well. If I have a whole scene that is consistent for noise, I'll do a range bounce and drop, that will drop a new clip on top of the group. Then, I can apply NR to that - again having any part of the original available instantly and in sync. Yes, you may get some latency on clip based plug-ins, but you can quickly slide it back, because Fairlight lets you see both the upper and lower layer waveforms at the same time. If I want to go out to a standalone NR app, I have the bounced clip available, do the processing, and drop it back - again with all originals available in sync in the timeline. I could do 20 different versions and just rotate through them by popping each up to the top layer. Great for the undecided client...
What's also cool about the clip based EQ is the it has gain of +20/-60dB, and a Q from .5 to 99 per band, plus Hi and Lo shelving. You also can store EQ settings globally on the system, so you can retrieve a specific EQ from a show 6 months ago, without loading that show or even having it on the drives. thumbsup

[mikevarela]

Are you guys saying that PT will actually nudge the audio in an audiosuite bounce of a NR plugin.

I'm amusing if this is true, ti means the audio in the region is nudged, not the region itself.

Also, by some estimates - how much?

[Garret]

Originally Posted by mikevarela

Are you guys saying that PT will actually nudge the audio in an audiosuite bounce of a NR plugin.

I'm amusing if this is true, ti means the audio in the region is nudged, not the region itself.

Also, by some estimates - how much?

No pt doesn't nudge the audiosuite of any plugin. Bill was talking about the fairlight which seems to, due to it being clip based. No need to worry about sync when using audiosuite

[Bill@AudioVision]

Originally Posted by Garret

No pt doesn't nudge the audiosuite of any plugin. Bill was talking about the fairlight which seems to, due to it being clip based. No need to worry about sync when using audiosuite

Correct, and only if you are using a rendering/non-real time plug-in. The Fairlight has adjustable latency compensation like PT for realtime plugs. But, the shift, for instance on TC DeNoise, is no more than 1/2 frame, so for most dialog uses, it's not a worry.