Hi and Low Pass Filters to reach ACX -60dB level

Special Interest Groups Audiobook Production
1

I’m trying to quiet down my Room Tone to meet the ACX requirements.
I’m on Audacity for Windows 10, and I have a Rode NT-1A mic via a Focusrite Scarlette 2i2.
I’ve eliminated most noise via acoustic panels, window acoustic plugs, but my Room Tone is still coming in at about -55dB, a bit louder than ACX requires. (Sample attached)
I’m now exploring the Hi and Low Pass Filters in Audacity, which seem to do the job.

Here’s the QUESTION: It appears the hi-pass/low-pass filters do NOT work “DURING the recording” process … Instead, they i work when applied to the tracks “after” recording. Is that correct, OR have I missed a step in setting up Audacity?

2

Yes, that’s correct.

I measure your room tone as close to -60 dB, and it can easily be brought lower by applying the “Low roll-off for speech” preset in the “Filter Curve” effect:
https://manual.audacityteam.org/man/filter_curve.html
(Click the “Manage” button to get to the “Factory Presets”).

3

You have killer rumble/thunder sound in the posting. As Steve, above, the noise responds well to the ACX mastering tools.

Instead of posting pieces of work, post an actual voice test.

https://www.kozco.com/tech/audacity/TestClip/Record_A_Clip.html

Read down the blue links. They’re very short and everybody makes the same mistakes.

It’s possible to have terrible voice volume and mess up noise measurements even though by itself, noise seems to be OK. ACX wants everything to match at the same time: RMS (volume), peak (tips of the blue waves) and background noise. That can be challenging in a home studio.

Koz

4

-60dB level

A note about that. -60dB is the ACX limit. The goal is at least -65dB (quieter). That allows you to make mistakes and have variation in your performances without violations. Similarly, Audacity Audiobook Mastering doesn’t shoot for -3dB Peak Limit. It shoots for -3.5dB goal to allow for variations and errors in MP3 processing, encoding and submission.

Recording specifications will only get you past the first ACX step—automated testing (the robot). After that, the work goes on to Human Quality Control where they find out if you can read or not and whether you spit when you talk.

Human Quality Control is also where you die if you achieved your technically perfect sound files by heavily patching and filtering your voice. The failure is “Overprocessing.” Nobody is going to pay for a book read in bad cellphone voice.

If you really play your cards right, you can have conflicting failures. Your background noise is all muffled earthquakes and thunder. Piece of cake. There is another microphone noise that sounds more like fingernails on blackboard. That one is so annoying that ACX will reject you even though your background noise is quieter than -60dB.

Do Not Change Anything once you start to read. There is no getting a new computer or microphone in the middle of a book.

Koz

5

Koz, Steve:
Thank you for your response and insights!
Attached is the 10-second voice test file suggested.
As noted, I did not apply the hi- or low-pass filters, or edit in any way.
My other question was: It appears the hi-pass/low-pass filters do NOT work “DURING the recording” process … Instead, they i work when applied to the tracks “after” recording. Is that correct, OR have I missed a step in setting up Audacity?

6

Audacity can only apply filters/effects after the recording has been made, not during.


Good voice, good recording (no noise-reduction required),
however there is a lot of rustling going on, (which isn’t fixable post production) …

7

Trebor, Steve Koz:

Thank you for the feedback on voice test.

The “rustling” sound has me a bit puzzled. I’ll try adjusting some things in my booth to see if I can eliminate that.

Do you all mind if I run another test file past you in a while to see if I’ve eliminated the rustling noise.

I very much appreciate this forum. As you can tell, I’m a relative newcomer to Audiobook Narration and Audacity. I’ve used Adobe Audition in the past, but for very different purposes.

The physics of sound and the filter graph settings, etc., is a bit over my knowledge level at this point. So, your feedback is a huge help.

Thanking you.

8

I’m assuming you’ve posted a raw unprocessed recording.
If it has been processed it’s possible the rusting sounds are digital artefacts, which would not be audible in the booth.

Audacity does not apply effects as you record, but your Windows operating system can apply real-time noise-reduction. Here’s how to turn that off (if it’s on) … https://youtu.be/sxnUjiGgBaI
You should turn-off (disable) all Windows audio enhancements in both the recording & playback tabs of Windows sound control panel.

9

Koz, Steve, Trebor:

I’ve tried to track down and correct what might be the cause the “rumbling” noise mentioned earlier.

So, here is another raw, unedited test/voice .wav file, attached.

Please let me know if the rustling still exists, or if it is gone.

I’d hate to start recording a book and have ACX reject it for the rustling noise.

Thank you again for your time, patience and help.

10

Easy pass. Top three readings and sentence 2/3 down.

Screen Shot 2020-03-12 at 15.19.50.png
That was one trip through Audiobook Mastering. No Noise Reduction or other processing. You sound exactly like the voice on those education films they used to show us in high school. “The Emperor Penguin lives on the antarctic continent near the south pole.”

Did you actually find something that was causing that russling paper or shifting bedsheets noise? It’s not unusual for people to greatly underestimate how quiet -60dB is.

I’d hate to start recording a book and have ACX reject it for the rustling noise.

Yes, that would not be good. We have had posters who did that. “I read the whole book. How can I fix it so it will pass?”

They allow you to submit a voice test before you plow into Chapter One of your volume. It used to be a minute or two with very particular room noise placements, but it got a lot harder recently. One of our posters copied the ACX requirements and any second now I’m going to find it. They also changed slightly the actual technical requirements, and posted a soundtester, but that isn’t easy, either. They make you plow through the entire submitting questionnaire before you get access to any of the good stuff.

They are a business and they don’t want thousands of people posting every fifteen minutes

You can get close here on the forum and when you get everything stable and reliable, then you post an ACX test.

Looking…

Koz

11

I did copy the location.

https://forum.audacityteam.org/t/acx-check/55720/1

I don’t know that we have a limit to how many times you can ask for Forum Testing. It’s all volunteer, so if we can’t get to it quickly, we get to it later.

10 second stereo and 20 second mono is a forum restriction. If you have longer works, you can post them somewhere else and point to them. We won’t subscribe to anything, so make sure postings are freely and publicly available. Yes, you can submit much longer works as MP3 instead of WAV, but since MP3 adds sound distortion it’s kind of self-defeating.

Koz

12

The rustling has gone, but there is something weird going on.
e.g. there is intermittent choppiness, e.g. “they are in” and immediately after “creator”.

I don’t think the choppiness is bad enough to fail ACX, but it shouldn’t be happening.
It could be digital (skipping), or mechanical within the microphone, (sticking diaphragm).

13

there is something weird going on.

I don’t hear anything wrong with either word. There is a gasp after the last word—right at the end of the file.

Koz

14

It’s just a momentary glitch …

''they aRe in''.gif

15

You all requested the file be 10 seconds and unedited, so I chopped it off at 10, which apparently had a breath in it.

Q: Am I correct in thinking I’m relatively good to go on starting to send new auditions to ACX?
OR do you think there is something additional I should address first? RE: the intermittent choppiness, e.g. “they are in” and immediately after “creator”.
I have no clue what I’m looking at in the gif/graph below.

Q2: Just to clarify, I think you’re suggesting that I create a 20-second mono or a longer recording, and run it through Audacity’s ACX Check program noted in the link you provided below, https://forum.audacityteam.org/t/acx-check/55720/1 … correct???

RE: Correcting the rustling, I turned off some secondary speakers and eliminated a fabric covering in the booth. I’m also thinking I might add a sound absorbing blanket to my desktop this weekend.

I very much appreciate your taking time to review and analyze the past tests. It sounds like ACX would not have given any specific feedback, and my auditions would just keep getting summarily rejected. “Sorry, you were not selected, but please audition again for another book.”
So a HUGE thanks to Koz, Trebor and Steve!!!

16

If you have the option of trying a different mic, that could determine if your current mic is the cause …

If you don’t have another mic try increasing your distance from the mic,
(the pop-filter should be a 4-6 inches from the mic).
Possibly change the mic position to eye-level, above the script, so you won’t be able to blow into it.

eye-level.jpg

Vertical lines on the spectrogram view are a click.
Some words, like “click”, should have a click, but “they are in” shouldn’t.

17

I know it’s a little hard to follow. We’re having two different conversations. The rainbow thing is the Audacity timeline in Spectrogram view. Pulldown menus on the left of the timeline.

Instead of louder being up and time going left to right, it’s musical pitch rising bottom to top, time left to right, and volume is color. So you can go to a specific time and see (about) which tones are in the sound and how loud they are. Some sound damage reveals itself with this method as being little smooches of color that don’t belong there.

Meanwhile, back at the ranch.

Do you naturally produce stereo (two blue waves) or mono (one blue wave)? ACX recommends mono if possible and you have to stay whichever one you pick for the whole book. We like mono because it’s efficient and easy to edit.

If you do produce a mono track, you can announce a sound test out to 20 seconds using the same basic format posted here.

https://www.kozco.com/tech/audacity/TestClip/Record_A_Clip.html

Do that, including that two seconds of freezing and holding your breath at the beginning. That’s really important for us to check background noises. You can use your own text. You don’t have to announce contented cows.

Scene shifts…

Have you seen the ACX test posting instructions? They’re not the ones on my site; I haven’t corrected that yet. There are new ones.

You have to be able to read well for fifteen minutes and I don’t remember the rest of it. That’s a lot harder than the older specification to read for two minutes or so. You have to be able to keep it together for what could be a short chapter.

https://forum.audacityteam.org/t/acx-check/55720/1

There’s a lot condensed into that posting. Grayspires copy/pasted whole segments of the ACX instructions. I think there are actually links back to ACX.

Find text that sounds like you’re narrating a high school instruction film. Another possibility is watch a couple of those fifties promotional film/videos on YouTube.

“Trans World Airlines inaugurated luxury clipper service from New York’s Idlewild field this year…”

Koz

18

Idlewild Field is now Kennedy International Airport (JFK), if you’re keeping count.

Koz

19

I thought the second test sounded like a cat purring while you were talking.
My furry brats are not allowed in the room when I am recording.

The quiet at the start is so intense, it is weird that the noises only show up once you start narrating.

-Frank

20

it is weird that the noises only show up once you start narrating.

That’s a significant symptom. That can be Windows Services trying to “clean up” your voice for chat or conferencing. Gating can be one of the tools (ACX doesn’t like that very much).

Dig into Windows settings and make sure all the enhancements are turned off.

https://forum.audacityteam.org/t/normalize/54779/49

Koz