Workflow for noise reduction in audiobook file?

[JeffOYB]

Hi... I'm prepping audiofiles for my first audiobook. I'm learning about the RMS and Peak and other ACX requirements -- I think I have a handle on those. I'm having more trouble with noise reduction. I've read here about leaving some breaths since they're natural. But I do have some mic hum. And I recorded in my house in a padded area of the basement, but kids were sometimes upstairs and some clomps and creaks got in. I don't mind having some little things here and there in the recording -- i'm not ashamed that it's amateur. ...But it wd be nice to remove mic hum and big clunks, breaths and mouth sounds that are obtrusive. While maybe just lowering some of the natural background yet not trying to totally remove it. Anyway, those are my goals.

When I google this subject I find a link to a Audition forum/community article titled "Workflow for noise reduction and speech volume." Sounds great but it seems to be gone -- part of the old forum, I'm guessing. Any new link?

I'm learning from watching a bunch of YouTube videos. So far I've explored:

*Adjusting the diagonal line on a Noise Gate in the Dynamics Processing. I see a Dynamics view and Settings -- i haven't figured out what the Settings do yet AT ALL. I see a lot of options but don't understand them. I find that lowering the left side of the line to the bottom at around the 40dB point gets rid of noise without cutting words. 

*Noise reduction using the capture noise print, along with its Advanced settings (lots of settings!) -- I'm reading here that doing multiple passes is a good idea. On YouTube the advisors all go for 100%... I've tested listening after a 10db reduction and 100db and can't tell what hurts the quality more! ...I still don't really understand dB even. In the world of sound it seems like settings with large numbers sometimes have less effect.

*Dynamics has AutoGate that seems to help when I set it at my Ave RMS ... but i dont know the other settings and it seems harsh.

*I changed mics a couple times during my recordings and I see that the resulting volume levels vary now and then. I'd like them to be uniform. It's been suggested in this Community that I could raise the yellow line or dial the dB and use my ears to do this. But when I increase volume on a quieter section once I get the "mid section" of the waveform to match the other sound there is a lot of "hairiness" to the wider markings. The waveform looks fuzzy compared to the other. Then there's Speech Leveling. Is there a downside to this function? After I use it the waveform does look totally uniform. My ear with my earbuds isn't good enough to tell me if it's done a bunch of something else as well. 

I've looked a bit at Compression and I see some YouTubers suggest it. (Julian Krause suggests some Tube Mod Comp.) But I'm not sure what it does or what it hurts or why I might want it. 

Lastly, sometimes my voice was a bit rough, sounding like I have a bit of a cold. Yikes... Sadly, it's early on. Later I get better skills and standards. I don't have time to re-record. (Heck I hardly have time for this processing.) Is there any way to improve overly low rough frog voice? Ha! Yeah, fat chance, but fingers crossed...

[SteveG_AudioMasters_]

'Workflow for noise reduction and speech volume' was actually about film production, and probably not quite as relevant as you might have hoped. Hum, you should be fixing at source - find out why it's happening and cure that - don't fix it afterwards. Yes, if you do vertical lines in Dynamics Processing you've made a noise gate - but a pretty vicious one that will be audible, because the background noise will bounce back every time a word is spoken. Mind you, that's pretty much always going to happen with a noise gate - the nature of the beast. I'm not going to explain everything about dB here - I'll just say that they are a way of measuring relative levels in a way that's similar to the way our ears work. Unless you specify a reference level, they are only relative units - IOW not like Volts and Amps. Any more than that, and it's going to get complicated.

All the people that claim that NR can be done in one go using huge numbers are DEAF!

Reading through the rest of it, I'd say that mostly these are things you should be fixing at source. It's much easier and quicker to do the processing necessary on a good recording than a poor one. As for your voice - well, don't drink anything other than clear liquids before you record, and keep sipping water during breaks. Other than that you may need a more sympathetic mic - but since we have no idea of what kit you have, it's difficult to be definitive about this.

[JeffOYB]

I was reading another thread about noise reduction and repeated passes at different FFT rates. I tried it a bit on a file and it didn't seem to work. I'll try it some more.

I have the files that I have and I'm going to try to clean them up. Any tips appreciated! Thanks.

[SteveG_AudioMasters_]

With NR, you have to bear in mind that all audio is different - and after a while you get a better handle on what it is you need to do that's going to be most effective on the particular file you've got. Sometimes there's not much point in using lower FFT sizes, sometimes there is. But whatever you do, doing it in three smaller passes, resampling each time, is invariably going to give you a smoother result with less underwater noises than a single pass taking a lot off.

[JeffOYB]

I spent another hour on the multi pass method and got no cumulative effect. Can you give more detail on how to do it? I read details about it in another thread and tried to emulate with no luck. ...Capture noise print, select all, 10% reduce, 10dB, fft 2000, apply. Then do it again at fft 4000, then 8000 then 16000... Do you also increase the % reduce and db? I've tried it keeping them the same and w increasing. No good luck either way. Not sure what i'm doing wrong...  I've tried advanced settings w Spectral at 0, 3, 5, 11% -- 0 seems fine. 

I've noticed that I've had to make multi passes even at 100% and 100db to get rid of warbles and loud breaths. 

[SteveG_AudioMasters_]

Like I said - all audio is different! Without a sample of it, I can't tell you what's going to be effective and what isn't. If there really isn't that much noise in the first place, then you won't be able to make a lot of significant changes anyway. Getting rid of all noise isn't desirable; it sounds completely unnatural, and this is another compromise you have to learn about - where do you stop?

But there's an important point to make here, in view of what you've said: Warbles and loud breaths aren't something you can use NR for. These are discrete sounds and not continuous, constant ones, and NR is only intended to work on the latter. You have to treat single noises on their own - often the Auto Heal favourite is the tool to go for - just highlight the noise and click on it. You may even have to do this more than once. Also it can be very effective if used in conjunction with the spectral display - if you can 'see' the noise you want to remove, then paint around it and Auto Heal just that.

[JeffOYB]

Any thoughts on Compression regarding narration for an audiobook?