Peak Levels

Hi, as a novice, I am having trouble getting all three specifications working at the same time - noise level, volume and peak levels. While I can manage noise levels and volume, it seems impossible to get peak levels to an average of -3db at the same time as the other two. Because Audible.com specifies ‘no more than -3db’ I was unsure if that meant they could be averaging -4db also. (The volume range is specific, so I wondered whether there was a lower limit to the -3db.) Apologies if this question is stupid! And a second question - if I get my volume levels okay, plus my noise floor is okay, but my average peak is either -2db or -4db, what is the least intrusive way of getting it in line? (I have discovered some effects produce a horrible tinny sound and want to use as few as possible.

And that in one paragraph is what’s been preventing us from writing the definitive WIKI on how to make an ACX audiobook file. It’s audio whack-a-mole.

If you got the noise and average loudness (RMS) to work, then you can cheat the peaks with Effect > Limiter

You don’t have to go nuts with “average peak calculation”, etc. Just run Normalize with a value of -3. No part of the show will exceed that value.

How did you shoot the work and what steps did you need to get all the values (except peak) to hit? Every fifth forum posting is “I’m a newbie and I want to record an audiobook.”

Generally, getting overall loudness (RMS) to work screws up noise and getting noise to work makes your voice trash. And then you whack the red mole and the blue mole pops up…

Koz

Thanks so much for your reply. It’s a relief to discover that I am not the only one with this problem…that everyone has this problem… I take from your answer that my peaks do not have to have an average of -3b - just so long as they do not surpass -3b?? Regards, Margaret.

Thanks so much for your reply. I did respond but it appears to have disappeared or something. Yes I am not sure what I did to get two out of three right - recorded quite loudly perhaps so I didn’t have to increase the volume too much. But I am much reassured (in one sense) by your reply, because now I realize I am not as stupid and incompetent as I thought. I take it from your reply, then, that so long as my peak volumes do not exceed -3db, it does’nt matter if the average isn’t EXACTLY - 3db? I daresay I will have further puzzled questions in due course… (Loved the red mole blue mole metaphor, by the way…)

Forum postings are moderated for the first few to prevent you trying to sell us something. We used to try and catch you in the act, but too many ads and SPAM went through. Now you have to wait for one of us to read it. It’s a pretty quick forum. It’s usually not a long wait.

Just out of curiosity, didn’t it say it was going to do that when you posted?

Koz

It’s my understanding they’re going to do a search for peaks over -3. If you violate that measurement too much or too often, they will bounce the submission.

Many people are trying to use simple USB microphones to produce audiobook work instead of going all out with a simple sound mixer and semi-pro microphone. USB mics make it really difficult to hit all the measurements at the same time. They were designed for personal podcasts and Skype which have a good deal looser technical specifications (no matter how many times they say “Pro” and “Studio” in the text).

The quality is usually good, but the noise levels are generally too high either from straight microphone hiss (fffffffff) or the semi-digital error of frying mosquitoes.

http://kozco.com/tech/audacity/clips/FryingMosquitoes3.wav

These are killers. It’s very difficult to bring your overall voice volume up (a very common client complaint) without boosting the background noise as well. The background noise specification is difficult to meet with a carefully designed home studio. It’s almost impossible to meet by slapping a USB microphone on the kitchen table.

Yes, we know all about Effect > Noise Removal. Noise Removal is a violin. You don’t just turn it on, say “Shazam,” and all the noise goes away. It has to be carefully tuned and adjusted for good sound without destroying your voice. Absolutely the opposite what everybody wants.

Next major version of Audacity will have an improved tool… and it will be called Noise Reduction.


Feel free to post samples of work as you go. I think you can get something like 10 seconds of mono voice in WAV format before the forum cuts you off. Give us a noise sample. Hold you breath and don’t move at all for one or two seconds, then go with the performance.

https://forum.audacityteam.org/t/how-to-post-an-audio-sample/29851/1

Lest we forget, this is all in addition to actual performance quality. This clip probably wouldn’t make it.

http://kozco.com/tech/audacity/clips/EchoSample.mp3

I’m sure she would be horrified I was using her ad as the Bad Example.

Koz

Thanks so much - that’s very helpful and I will certainly take you up on the post-a-sample later today or tomorrow. I need to know the truth about the quality of my efforts :slight_smile: Cheers, Margaret.

I need to know the truth about the quality of my efforts

You will get our opinion. Some of the posters come very close (as near as we can tell) to ACX Audiobook compliance.

I know (or strongly suspect) that your submission goes through automatic compliance first to weed out obvious failures, but then goes to human testing and that takes a good deal longer. That’s when you get a personalized review of the shortcomings and can be very valuable.

Between a refusal notice and posting a sample of the work, we can usually guide you through repairs. Sometimes not. If you have bad equipment or bad environment, you’re stuck with major changes. We have at least two readers who have to read around street traffic or railway noises.

You probably should not crank through a major work before testing. I understand ACX will accept a 15 minute test submission, but evaluation can take weeks. We can usually get you close, faster.

Everybody wants to read a novel.

Koz

Well, here it is - the first line of Jane Austen’s Pride and Prejudice. It is within the volume limit, within the peak limit and the floor limit is okay. So technically it is okay,as far as I am aware. But is it okay? Or do I have to throw away the kitchen table and save up for a better one :slight_smile: ???

“It was the best of times, it was the worst of times.”

Sorry. Wrong one.

Do you remember the Noise Reduction setting you used (top number)? It was probably too high whatever it was. The silence segment at the top is too silent and the spaces between words and at natural gaps is too deep. If I advance the playback volume, the silences contain odd bubbling, gargling sounds typical of digital processing. So while your actual words are lovely, everything else screams Digital Voice.

I have no doubt the clip would pass first echelon conformance, I don’t think it would pass the first time a human listened to it.

My opinion. Remember you asked for this.

We’ll see what the other elves think.

Is Sydney Carton in this one? I liked Sydney Carton.

Koz

Who is Syndey Carton :slight_smile: ? My noise reduction settings were 20 for noise reduction, -5.20 for sensitivity and 369 for sensitivity smoothing. Thanks for the feedback, it is exactly what I wanted - the truth. Would it suffice, do you think, to just go easier on the noise reduction, or is the problem deeper than that? Cheers, Margaret.

I think you’ve got a nice narration voice.

It sounds a little muffled though.

I would give it a stronger foundation.
Try adjusting the base for once. Either “Bass&Treble” with 6 dB base boost (none for the treble).
You can go as high as 18 dB (or use the effect 3 times).
use the amplify effect afterwards. If it proposes “1 dB”, enter “-2” dB to reach the -3 dB head room.
The volume (Rms level) should be better as well, around -20 dB.

It is clear that the equalizer would allow a better fine adjustment.
Raise e.g the sliders 100 125 and 160 Hz by a considerable dB amount (200 Hz has enough power already, I’d say).

There are some other gaps in the spectrum that would profit from a little boost, in order to add presence to your voice.

Who is Syndey Carton?

“A Tale of Two Cities” character.

20 for noise reduction

That’s right on the edge. I pick 24dB Reduction as the signal that you need to adjust your room or system to get rid of noises instead of trying to do it in post production software.

Do you have the original, raw, unpatched WAV clip or Project? If you say no, that’s one demerit. Original capture work is golden and should be saved as high quality WAV someplace safe. If you don’t do that and something evil happens during post production, you have no show. Grab some coffee and read the whole thing again.

“Koz, do you still have that shoot we did last month?”
Of course. Want me to post it for you?

Anyway, yes. It would be good to hear what the original shoot sounded like. It’s possible that original work doesn’t have some of the sound quality problems that Robert doesn’t like. Noise Reduction is not without its problems.

Koz

I was having problems with peak levels. I ended up just turning my mic input level down in Audacity. So far it seems to be working, but I’m not sure how it would sound for others when I send in my clips. I’m preparing to record a children’s book, but I want to make sure everything is of good quality first. Can I send in some sample clips, have someone listen to them, and give advice?