Is there a way to find all peaks over a set level [-3db]?

Special Interest Groups Audiobook Production
1

So, I got through recording my entire book (yay!), following the Honorable Koz’s method of EQ>Normalize to 3.2>Compress>Normalize to 3.2…

The files met ACX requirements. Life was good. Then, by happenstance I recorded the last chapter speaking slightly closer to the microphone. It sounded noticeably better. Like someone was actually there, instead of listening to a recorded voice, if that makes any sense…

BEFORE, my raw tracks were consistently around -9db before normalizing. The occasional exclamation or stress would get up to -5db or so, and then the normalizing would pull everything up.

NOW, my raw tracks are dancing around -3db in every sentence. The waveform looks more dynamic. Almost every other sentence, though, I go over -3.2db.

If I were to just blindly normalize to -3.2db, I would likely be pushing everything down a couple decibels, as I surely have a -1db peak in there somewhere. That’s no good - I want to keep the quieter portions as is.

So, I’ve been manually finding peaks over -3.2db (using the ACX Check plugin), and normalizing only that word to keep it under the threshold. The result sounds great, but holy moly is it time consuming.

My question then, is if there is a better way to do this? I suppose compression with a -4db threshold and 3:1 compression would work in theory? But I’m afraid to try that and then compress AGAIN later. Side note - I actually met ACX standards before compression and normalization on one of the tracks.

Related, is there a way to have Audacity highlight or somehow show you all instances of the sound going over a set amount? This would save me tons of time.

Or, am I terribly wrong with all of this and I should just shut up and go home?

2

Just to be clear, the 3.2dB value while the show is high quality WAV is so when you do convert to MP3, the final MP3 doesn’t go over 3.0dB. MP3’s job is to reduce the file size while making the show management and manipulation inaudible. Full Stop. Nowhere is it written that the files have have specific sound peaks or blue waves. Most people have no idea what those are. The conversion of WAV to MP3 is a favorite place to get into trouble.

“How come my MP3 is higher…”

Music and complex sounds are made up of many different sounds, overtones and harmonics. They don’t all just add up. The job of some of them is to subtract. If you’re having trouble sleeping, we’ll explain how Fourier Transform works. “The algebraic summation of the harmonic…zzzzzzzzzzz”

If MP3 gets it in its craw to delete one of those degenerative sounds because it will make the sound file smaller, the volume may actually go up slightly.

Anyway back to real life. You might like the limiter tools. Their job is to suppress peaks while more or less not affecting anything else.

And…that sound you hear is me digging myself a hole. I think the upgraded Effect > Limiter is only available in Audacity 2.1.1 which I don’t have that on this machine…

The on-line instructions will tell you what to do.
http://manual.audacityteam.org/o/man/limiter.html


It is concerning that you got slightly, almost insignificantly louder and yet you sound so much better. Are you suffering from The Music Store Syndrome? The loudest speaker always sells best.

Even with a good quiet room you can get a significant theatrical change in the last bunch of inches as you approach the microphone. I’m a bass. I once performed a passable woman when that character’s turn came by leaning very closely into the microphone and “performing” the woman’s part. The producer left it in the show.

The down side is critical spacing. The closer you get, the more you’ll need a clamp to keep your head in the right place. The slightest movement will change the character and volume of the performance.

But past that, whatever works. If you like the new sound, then by all means do performances like that.

Remember how to do it. Nothing like being away for several weeks and not remembering how you did it. Almost as much fun as greatly improving the performance…on the last chapter.

Koz

3

and a direct answer to your question…

Yes, you can use “Silence Finder with the silence threshold set to the level that you want to mark”; http://manual.audacityteam.org/o/man/silence_finder_and_sound_finder.html#silence

4

Boom.

Thank you Steve! That’s going to save me a lot of squinting looking for the slightly higher points.

And Koz, fully understood on the risks of getting very close to the mic. I can even see, within the one show, how periods after I come back from a break are a little off until I settle in to the right position again.

Basically, I “figured it out” for Chapters 24 and 25. Most of the previous chapters were salvageable by simply tuning down some peaks and re-normalizing, to bring up the RMS of the show about 1db, which makes it sound much fuller. Not quite as full as the last chapters, but close enough someone not listening maniacally like me might not notice. Chapters 1-3, though, were an unmitigated disaster both aurally and performance-wise, so I am re-recording those with this closer-position method.

I’m going to see how the shift from Chapter 3 to Chapter 4 sounds before I commit to (gulp) redoing them all.

I’ll post the result here, actually, as I’m open to tricks to tweak the tracks to sound like the “better” ones… maybe something with an EQ adjustment?

5

It’s perfectly normal to get to the end of the first book, listen to the first couple of chapters and start over.

It’s nice that you think applying a little Eq is going to solve your problems, but most people reading at great length for a long time get better at acting. We don’t have a filter for that.

Koz

6

C’mon, Koz… I may not know much but I know I can’t EQ out a bad reading :wink:

Attached is what I’m speaking of… Obviously I’ve improved my reading cadence and such, but what I’m referring to is how Ch3 sounds, I don’t know… fuller, better. Their RMS values are similar, but I feel like Ch4 sounds more echo-y or hollow, because the raw recording was at a lower db level and it had to be pulled up.

What I was asking (hoping) about was if a file that sounds like Ch4 can be EQd and played with to sound like Ch3? Along the process of my book I feel there’s some chapters where the reading is up to par, but the sound isn’t quite as good as what I’m doing now - if I can salvage those without re-recording, that’s great.


I’m guessing there’s not much that can be done, and I’ll need to re-do it to capture a stronger signal… but it’s a Hail Mary…

7

I’m not able to listen to the audio samples right now, but re. making the level more even:

  1. It does not need to be, and should not be dead flat - all one level. “Dynamics” are good (add interest and expression) when they are intentional and not too OTT.
  2. A “Compressor” effect and/or a “Limiter” does automatically what you are doing manually. If you just need to knock some brief peaks down by 2 or 3 dB then I’d suggest that you (after making a backup copy) try the Audacity Limiter http://manual.audacityteam.org/o/man/limiter.html
8

To that end, Steve…if I find that the track already meets ACX standard before compressing, should I just stop there? Or should I compress and normalize anyway? I can see an argument for both.

9

Also, after reading about the Limiter…a question…what is the difference between soft clipping and a soft limiter? Both appear to gently push the wave below the limit

10

Yes. If you compress too much it will start to sound tediously loud. Keep some dynamics in.

“Peaks” means both upward and downward pointing peaks.

“Clipping” chops the peaks off.
“Soft clipping” rounds the corners a bit, so is less harsh than “hard” clipping.
Clipping always creates distortion, so should not be used unless you actually want to add distortion.

“Limiting” is like a very fast “compressor” effect. It reduces the gain when the amplitude reaches the threshold level so as to limit how high the waveform goes.
“Hard” limiting has a very distinct “threshold” level. Audio below that level is completely unaffected, and audio above that level is limited to the specified maximum level.
“Soft” limiting is more progressive in that it reduces the gain a little as the waveform approaches the threshold level, The gain reduction increases progressively . In compressor effects, this is referred to as a"soft knee". This article about compressors explains the concept of hard and soft knee: http://www.homestudiocorner.com/what-is-knee-on-a-compressor/
True “limiting” (as opposed to “clipping”) produces very little distortion, so is generally better for high quality audio (unless you actually want to add some distortion), This is true of both “hard” and “soft” limiting. The difference in sound between hard limiting and soft limiting is quite subtle and not easy to describe - best that you experiment :wink:

11

if I find that the track already meets ACX standard before compressing, should I just stop there?

The AudioBook Goals are that your voice sound pleasant (and real), the volume be generally even chapter-to-chapter and you make the ACX Robot happy.

Have you listened to the whole thing all the way through?

Koz

12

Remember the original question? You could also Effect > Amplify the track by 3.2dB (allow clipping). If Show Clipping is turned on, all your 3.0dB points will turn red. When you’re done managing the sound, Amplify by -3.2 to bring everything back down.

Audacity works in an internal format that doesn’t clip.

Koz

13

Not yet. I did the whole book in order, 1-25, and by 24 realized it sounded better by hitting ~3db in my raw files than the ~6db I was hitting before. So I did 24 & 25 that way, went back and listened to 1 & 2, and they were terrible by comparison - although to be honest it was mostly my reading that made it atrocious.

So I re-recorded 1, 2, and 3 so far, as those were the first three I did, on the same day, my first day. The next chunk of chapters sounds a little better.

Basically, I feel like my reading improved considerably through the first ten chapters, then I was pretty much ok after that. The sound quality was crap on 1-2, ok for 3-23, and great for 24-25.

I’m now re-recording chapter by chapter to see if the dropoff from one to the next is small enough to stop re-recording. Something tells me it’ll drive me nuts, though, and I’ll end up redoing them all.

14

Possibly dumb question - I’m running 2.1.0, and I don’t see “Limiter” as an effect option… I see “hard limiter” further down, but the interface is different?

15

The “Limiter” effect is new in Audacity 2.1.1 (get it here: http://web.audacityteam.org/download/)
I wouldn’t recommend the old “Hard Limiter” or this type of job as that is a “soft clipping” type limiter.

16

Words like "soft, “gentle,” and “gradual” are good because the After sound is most likely to resemble the Before sound. One of my processing jokes: “I did all that processing and my show still sounds like me.” ACX looks for voices that stop sounding human.

One problem with Audacity 2.1.1 is that it doesn’t remember Equalization Effects from one application to the next. It was a programming oversight. 2.1.0 does. We assume 2.1.2 will, too.

by hitting ~3db in my raw files than the ~6db I was hitting before.

3dB is a barely audible change, either up or down. If you’re getting a massive quality different by getting slightly closer to the microphone, then you are probably triggering proximity effects or other processing in the microphone system. You are also in much higher danger of overload and clipping that high.

-6dB was chosen as the happy medium between performances so low that they compete with the natural noise levels (fffff), and so high that any emphasis or theatrical stress on words may damage the sentence through overload.

It is totally a juggling act. If you can manage presenting that high for long periods without wandering off volume, then that’s the way to go. There isn’t isn’t a gold-plated rulebook of Do This and Do That. But there are conditions likely to cause problems and conditions and settings that many people have found helpful.

Oh, and no, Audacity doesn’t apply tools, effects or filters in real time, so you can’t automatically process your voice (although there are mixers and MicPres which can do that).

I feel like my reading improved considerably through the first ten chapters, then I was pretty much ok after that.

Yes. But the voice and presentation quality is usually so much better people just go back and read the early ones again.

After you settle on your final environment and presentation technique, things will be a lot better on the second and third books. You might save your -3 technique for the second book.

Koz

17

Thank you all, But from the whole conversation I learned that Audacity does not have any actual tool that can shows you the highest peak or peaks of a Track. Right ??

18

As you’ve started a new topic, let’s discus here: https://forum.audacityteam.org/t/how-can-i-find-the-track-peak-or-peaks/54189/1