Need help with audiobook noise levels

the macro will apply whatever the last equalization curve used was

Or not. Remember the Equalization difficulty where it will remember the last curve changed, not used. Some Audacity versions have this problem. Also, there is some doubt whether calling a specific curve in a Macro will work either, because of another difficulty where Equalization seems to have no idea which curve is mounted.

It is a puzzlement and that’s why I have not advocated a Mastering Macro for wide use. If you use the tools manually, you can see which curve presents right in front of you and make corrections if needed.

Koz

This is exactly what had happened to me. I took that photo without realizing that it had defaulted to the unnamed preset again- which it keeps doing.

Is that picture of my sound waves? Meaning I have a DC offset I need to address? Or you are showing me an example. Where can I learn more about this stuff without scrolling through hundreds of forums to try to figure out what a DC offset is.

What a terrific idea [writing that down]

Uh huh. I’m not so brilliant late at night.

I ran ACX Mastering (you can use your Macro) and then I dumped it directly into DeEsser at the posted values. Full stop. I don’t think I used any Noise Reduction in your corrected clip. Now that I have a handle on how DeEsser works, you might try this protocol on a longer piece and see if sounds OK. Pay attention to your mouth ticks and see if maybe you can just ignore them now that they’re not little ice picks in your ear.

I went through the process I described yesterday, using the De-Esser at the end and it sounded great. I had already removed most clickey noises though. I ran the De-clicker at the beginning with your settings and that gets rid of most of them. The Noise reduction of the beast settings helped the most with the room noise- that was great! I was just scared to run processing before. It’s not that I didn’t know what the tools were, more that I was seeking validation that it was okay to use them and still submit to ACX

ACX tends to regard any natural sounds as, well, natural and they don’t object to them. I think people who go through and remove breathing noises are wasting their time…with one exception. There was one borderline asthmatic trying to read. That may not have been the best match. That was painful to listen to.

I didn’t remove anything but weird breaths. I have a tendency to go as long as I can without breathing (Im a singer, so this is kind of ingrained) and take big giant breaths. I’m better about it now, but my earlier recordings had some weird breaths in them. I leave the natural ones in when I can. Unless there is a big mouth noise at the beginning of it.
—Heres’s a question for you, initially I was using noise reduction on breaths that sounded unnaturally loud- to try to make them less noticeable but still there, so it sounded more natural. Thoughts on that, obviously it wouldn’t work for EVERYTHING, but just surgically here and there. I think the biggest problem I ran into was that it was TOO silent on the room tone around the breaths when I did that.


You are saving raw readings as WAV files, right?

Yes! I save them at the end of recording as a “pre-edit” then I save them after cutting and pasting as a “pre-mastering” then I save as a wav file and a mp3 after mastering.

I don’t know what a non-symmetrical voice is lol- I’m guessing I have one, is it because I have a lot of variation between loud and soft?

See the blue waves have an up and down middle at Zero (on the left). Some voices have uppy waves that don’t match the downy waves. They’re non-symmetrical. That has to do with mouth construction, sinus size, breath control, etc. etc. Some performers are convinced the world will end because of this effect. Probably not. I’ve never seen the tools get confused by that.

Still confused by this…don’t my uppy waves match my downy waves? (are those the technical terms lol). I’m not concerned about it, I’m just trying to understand

Surprise. You picked another display variation. You selected Waveform (dB) with the drop-down menu to the left of the blue waves. Audacity defaults to Waveform, which is percent rather than dB.

Like this:





0 percent in the middle (dead silence), 50% half-way up and 100% all the way up.

This is why in some difficult problems we recommend you post a screen capture or grab of your work. So we’re both singing from the same hymnal.

Both waves tell you stuff, but in Waveform (dB), the critical area of the waves is in the top 1/16 inch of the screen. Rough to see well.

In Waveform, it’s easy to see when your blue wave tips just occasionally touch 50%. like in the above picture.

Yes it was an awkward choice to have the bouncing sound meter and the blue waves measure the same thing but look different. That’s a discussion for the developers.


While we’re juggling tools for the best possible product, we should keep in mind it is a product. The more cheaply you can make it and be acceptable, the more profit you get. It’s not valuable to include filters and corrections without a very firm reason. You will be wasting time by applying a long list of corrections for every chapter for a whole book. Also note all the chapters should match, so you can’t stop in the middle and say you’re going to change the protocol.

I would reference the reader who assured us he was going to carefully correct every word for a whole book. He’ll be doing that for a very long time and the chances of success are very low because he will never make it sound natural.

That’s why it was welcome news when you produced the Mastering Macro. One tool instead of three. It was very unfortunate you ran into an Audacity tool difficulty.


I would correct a raw reading without picking apart each and every tick and see if DeEsser takes care of it for you. That would get you down to Raw Editing, ACX Mastering, DeEsser and then Noise Reduction.

Another valid variation is Raw Editing, Noise Reduction, ACX Mastering, and then DeEsser.

Koz

Do you NOT recommend the declicker? That seemed to have the greatest benefit to me? Ive been running that before I go through and edit.

I’m about to work on a file that has been unedited- just remove the “toddler wanting to be fed” noises, since I don’t have a cat :wink: And then I will run mastering Noise control and DeEsser and post it.

I’m still not sure why my voice doesn’t look asymmetrical in Waveform DB but does in Waveform. Thats confusing. The only reason I use Waveform DB is because some tutorial I must have seen a month ago told me to do it. And I can see what I’m doing better. I can’t see what I’m doing at all in Waveform, no matter how far I zoom in?
Screen Shot 2019-03-27 at 6.06.39 PM.jpg

In studying everything you wrote yet again, I’m wondering if I’m just too far from the microphone? It appears that my percent is just a little over .10?

Do you NOT recommend the declicker?

I don’t do anything with DeClicker. I’ve never used it. I do know that mouth noises are boosted by some microphones because it “sounds more professional.” Not in my book, but I’m not making microphones.

I wondered if one good pass through DeEsser can solve multiple problems. Are uncompensated clicks and pops quieter after DeEssing?

This is your show. If you like the way it sounds, then you win.

I’m wondering if I’m just too far from the microphone?

If you are within shaka range then the spacing is just about right. However you may not be digitally loud enough. Isn’t there a volume control on your microphone interface? The object is to announce normally and advance the volume control until your Audacity bouncing sound meter reaches occasionally up to -6dB—like the illustration. That works out to 50% on the blue waves measured in percent.

That’s why it can be misleading to use Waveform (dB), You can be way off volume and still think you’re doing fine because of the mass of blue in front of you. Most of that blue can’t be heard, only the loudest about 24dB. That’s why the Audacity default waves are not dB.

This might not be a bad place for a Photoshop® graphic…


Your interface is the Ai-1, right? What do the instructions say about that green light on the upper left? The web page doesn’t say anything.

Does it flash green when you speak? If you advance the volume control, can you speak and make it turn red?

If you’re recording in Audacity what happens to the blue waves and the sound meter when you do that?

Koz

Your interface is the Ai-1, right? What do the instructions say about that green light on the upper left? The web page doesn’t say anything.

Does it flash green when you speak? If you advance the volume control, can you speak and make it turn red?

If you’re recording in Audacity what happens to the blue waves and the sound meter when you do that?

Koz

Yes, I have the Ai-1, and there are 2 dials, one affects the gain and one the volume on the headphones. I adjusted it until it was green when I speak, not going into red. It’s sitting at about 2:00 or about 3/4 of the way turned. I have that pop filter on that comes with it, and I can advance it until it turns red when I speak. Should I be closer to that? I’m literally measuring a “Shaka” (Shakra) away from the actual microphone, not the pop filter.


Also the Declicker removes the mouth click noises better than the DeEsser

I edited your post to remove the odd quote marks.

Also the Declicker removes the mouth click noises better than the DeEsser

So DeClicker it is, or both in this case.

I’m literally measuring a “Shaka” (Shakra) away from the actual microphone

Chakra is for balancing your spiritual energy, Shaka is a Hawaiian greeting. Chakras are terrible for measuring linear distance.

You appear to be doing everything right.

An experiment. Announce one of those 20 second tests, but turn the volume knob up from what seems to be normal for you. Announce as you go. “This is normal.” “I got the first flash of red light right there.” 'The red light is getting brighter and brighter." “The knob is now all the way up, One, Two Three, Four.”

I eventually found the instructions for the interface. They said something about the digital conversion I didn’t understand. We should resolve this before I try telling people how to use this interface. I don’t entirely trust lights that try to “help me.”

Koz

Heres’s a question for you, initially I was using noise reduction on breaths that sounded unnaturally loud- to try to make them less noticeable but still there, so it sounded more natural.

There’s the Envelope Tool (two vertical black arrows and bent line). That lets you draw volume corrections right on the screen.



I went nuts with control points to show you it was possible to get a very graceful and gradual change. You can put the control points anywhere you wish. Drag a control point off the screen to get rid of it. Don’t forget to go back to the “I-Beam Tool” when you’re done.

Yes, sudden Blackness Of Space changes in background volume are not welcome.

Koz

Yes, I’ve mastered the envelope tool :slight_smile: It’s just VERY TIME CONSUMING, initially I was hoping for a shortcut, and doing small sections of noise reduction seemed to do it much more quickly, but alas too abruptly. There are shortcuts to somethings, and NOT to others it appears :wink:

I have recorded a test, it appears when the “gain” knob is turned to about 2:00 I get a nice even green light the whole time. That’s what I thought I was shooting for- when I’m about 4:30 or so is when it seems to be mostly yellow and green. Then about 5:30 ish is where its mostly red. I think I just recorded everything at too low of a volume! I couldn’t find any information on what level I should have the gain turned to, and it seemed like having it stay mostly green was the goal, so I just turned it to where it stayed green the whole time (or MOST of the time) I was narrating.

And I totally see what you mean about the asymmetrical sound waves. That’s interesting. Seems to be because I have a lower pitched voice, but still have some brighter tones to it.

Also. Thank you! Not just for answering my questions, for which I am supremely grateful, but for also taking the time to answer so many other questions. Searching these forums has given me a wealth of knowledge. Hopefully time will give me the experience to use it as well.

I was hoping for a shortcut

You were hoping for a digital shortcut. How about you turn your head and gasp off-mic.

You learned to do that breath trick as a live performer, maybe you need to learn a different technique for audiobook reading.

I think I just recorded everything at too low of a volume!

And there’s no good way to know until you produce that test sound clip. There’s a very good chance that high volume sound damage may happen in the interface at a lower volume than it does in Audacity. It would be super good to know that.

One of the obsessive engineer tricks is to not only know how my equipment works, but how it acts when it’s broken. That’s the sort of thing you don’t find in the operator’s guide or instruction book. What does the Audacity display look like when the Rode interface is flashing yellow? Red?

Koz

a Photoshop® graphic…

Yes, you can use Waveform (dB) to watch the work, but that may not be as useful as you think.

This is a made-up performance with about the right volumes in Waveform.



It shows the tips of the waves at about 50% (which is -6dB if you’re counting) and is just about perfect.

This is the same performance in Waveform (dB).



All the critical parts of the wave are crammed into the red zones, which we all know is for immediate loading and unloading of passengers only.

The red areas are trying to display overload (clipping), peak recording volume, -3dB Peak ACX and the ACX RMS loudness range. All those important values are smashed into the tiny red areas in Waveform (dB).

It gets worse. If you’re doing this as a way to see your noise volume down in the -60 range, that may still not be very useful without extending the graphic quieter than -60dB which will make the red areas even smaller.

I know people who use Waveform dB for everything, but their work is existing recordings which are a lot better behaved than live speech.

Koz


Apparently my attachments aren’t working? Let me try this again

Attachments may fail for being digitally too large or in a format the forum may not accept. It can also fail if you forgot to send it. Of course, that never happens to me.

Koz

Terrific.

The interface does perfectly expected things. The borderline between green-ish and yellow corresponds to our recommendation for normal recording volume—tips of peaks roughly around 50%. Heavy yellow is OK, but you are spittin’ distance of possible sound distortion. Any theatrical emphasis during that passage and you’re dead.

Make sure Audacity View > Show Clipping is turned on. Note in my attachment the thin red lines at the “Red Light” announcement. That is distortion where the digital system stopped following your voice because it was too loud. If you do enough of that, the damage is harsh, obnoxious and permanent.


Screen Shot 2019-03-28 at 10.44.51.png
So you can keep one eye on the interface and pretty much see how you’re doing. Also, very highly recommended is listen to yourself on the interface headphones connection. People’s voices tend to even out when they can hear themselves in real time as they perform.





Oh, the instructions? They suggest a lot more damage than actually occurs, no doubt to keep New Users out of trouble. In fact, the interface works exactly correctly and I have no trouble recommending it. This was an insanely valuable exercise. Thanks!

Koz

GOOD! I will turn the knob up a bit the next time I record, obviously not for this book since I’m done recording. I have “show clipping” turned on, but since I had the gain set so low it has never spiked red until this particular exercise.

So just to be sure I understand- where I had the knob turned at about 10 seconds (about 4:00 on my dial) would be the recommended setting- Mostly green with just spikes into yellow- and no red.

And yes, I learned very fast that I needed good quality headphones, and purchased some right at the beginning. When I am recording I keep them plugged into the interface.

Thanks for all of your help!

Mostly green with just spikes into yellow- and no red.

That should work, yes.

Koz