Am I Doing This Right?

I’m using Audacity 2.1.0 to record an audiobook (fiction). I already saw there’s an update, so I’ll apply that. I just found this forum so I want to run my method by others to see if there’s a way to do this better or more efficiently. It takes a ton of time! I’m using a ATR2100-USB Audio-Technica mic plugged into my Asus/Windows 10 laptop and recording in a walk-in closet in the central part of my house, using a cardboard box lined with noise-protecting foam to encase the mic. I have the foam sock over the mic to shield breaths, etc.

  1. set recording channels to mono.
  2. Set project rate to 44100Hz.
  3. After recording, I go through and remove any parts that I repeated due to mistakes, then I take these steps for each track. Select a second of clean, room tone noise and do Effect->Noise Reduction and Get Noise Profile.
  4. Then highlight the whole track and do Effect->Noise Reduction and use Noise Reduction 48db Sensitivity 0 db Frequency Smoothing 150 and apply
  5. Effect->Compress Dynamics 1.2.6 and Compress Ratio .75 Compression Hardness .6 Floor -24 Noise Gate Falloff 4 Maximum Amplitude .99.
  6. Then Analyze->RMS Calculate to ensure it’s between -20.1db and -19.4 db. If not, undo the last step and adjust the compression ratio up or down and repeat.
  7. Run the hard limiter at -3.
  8. Then I grab a second of clean room tone again and go through the track and replace any of the pauses between words or sentences with it, adjusting to go back to the original length after copying over it with the second of clean noise. This part takes forever. It’s to remove any remaining breaths, lip smacks, etc. I’d like to be able to reduce this time. I actually repeat this with a much shorter length for very small intervals between words where I can detect any of that excess noise.
  9. When that’s done, I listen again to make sure it sounds okay, then export as MP3 at 192kbps and upload to ACX.

Any tips as to how to do this better or reduce the long hours it takes, especially for step 8? Any help is appreciated. I’m already most of the way through this book but have two more to do after it.

Attaching a sound sample in case anyone can listen and provide feedback. Will I get slaughtered by discerning audiobook listeners? I’m also not sure if I had the foam sock on the mic for this first recording.

I forgot to add that in the well-stuffed walk-in closet, it’s just me, the mic, foam mic insulating box, and the ipad that I’m reading from, and my audio-technica over the ears active noise canceling headphones (plugged into the port on the back of my mic). I put the laptop on a stand outside the closed door and run the USB cable to it under the door. Closet is also carpeted and I turn off the house AC and everything else I can find.

There’s a free DeClicker plugin for Audacity here which will automatically remove lip-smacking …
Suggested settings to remove mouth clicks.png

The noise-reduction is native to Audacity, the noise gate is a plugin from here.

Any tips as to how to do this better or reduce the long hours it takes, especially for step 8? Any help is appreciated. I’m already most of the way through this book

You may already be partially hosed. Changing the process or sound quality in the middle of a book is highly not recommended. ACX is very clear they want good chapter to chapter matching. No changes to take you out of the story flow.

I’ll listen in a minute. You’re using a version of the older technique and I wondered why more people didn’t use Chris’s Compressor to take the volume variations out of the work. But you’re right. That’s a lot of drudge per minute of performance.

The new recommended mastering suite has custom-written tools and is way shorter and faster, but it does assume you’re not going to go in word by word and correct it. Did somebody complain about that, or are you just being compulsive?

Audiobook Mastering Ver 4.

If you have a pleasant microphone and good volume, it boils down to three steps: Rumble filter, set RMS (Loudness), beat back any high peaks. Go make tea. Four, including the tea. We can produce an ACX Technical compliant work very quickly. If you have noise, the process goes into extra innings and if you make mouth noises (holding up hand), you may need some of the vocal quality filters.

Noise Reduction 48db Sensitivity 0 db Frequency Smoothing 150

48dB Noise Reduction is stunningly high. The new suite recommends 12 tops, more like 9 or 6.

If you have a half-minute, burn us a forum test voice clip. We want the raw work. No processing.

Record > Stop (cut it down to 20 seconds) > Export.

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

Read the back of the Corn Flakes box, anything, but do it in your normal voice. One poster read the instructions to produce a test clip.

I am not solving your current problem. You may be able to use this newer technique on the second book.

Koz

I’m using Audacity 2.1.0 to record an audiobook (fiction).

I just saw that. You’re stuck again. Audacity 2.1.3 is required for some of the fancier new tools, but I would not change anything in your current reading. Little sign over my desk: “If you find something that works, hold onto it with white knuckles.”

Did you submit a test to ACX before you started reading your book?

Just wondered.

Koz

I could listen to a book in that voice. If you used the sock, you may not need as much correction—or any at all.

Sometimes excessive Noise Reduction causes harsh peaking, essing and sibilance problems. So solving air conditioner fan noise problems turns out to solve voice quality problems down the line. That’s what was so darn much fun making the new audiobook suite. Where do you even start?

Koz

Thanks Koz, your contributions here are amazing. I wish I would have found this before I recorded! I’ll post a raw voice sample tomorrow. I did submit a sample to ACX but haven’t heard back from them in weeks, so I went ahead, regrettably. My process is so tedious, I think I can actually re-record in less time rather than do the rest of the book this way. I’ve already been reading the threads here about the new tools and installed them, so I’m ready to go. That’s what my weekend will consist of! Do you recommend also using the declipping tool mentioned above for lip smacks? I have been able to improve my technique to lessen those (and using lip balm, etc) and breaths.

But what do you do if you are toward the end of a long chapter and suddenly make a mistake in a sentence, say pronounce a word wrong? My previous technique would be to just re-read the sentence, and then when in the edit/review phase just clip out the bad one. Is that still the way to go or just re-start the whole track and chapter from the start if a mistake happens?

My process is so tedious, I think I can actually re-record in less time rather than do the rest of the book this way.

That’s what wearing the Producer’s hat is like. Isn’t this fun?

Do post the test clip.

re-read the sentence, and then when in the edit/review phase just clip out the bad one.

That’s how I do it. You can leave a little of a gap so you have room to cut and still make the sentences sound natural. You can also post a label right there with a keyboard shortcut if you do that in the middle of a long chapter. If you do it quickly like that, you can maintain your theatrical rhythm and expression much better.

Windows would be Control-B to set a label… I think.

http://manual.audacityteam.org/man/label_tracks.html

Of course, just as I said that, the labels on my machine failed. ~!@#$%^

Koz

Here’s my raw sample. I noticed when exporting this that there seems to be a new setting in the mp3 export, a box that says ‘force to mono’ or something like that. I assume that should be checked going forward? Thanks for all the help, let me know if there are any changes I should make to the environment or settings based on the raw sample attached. I’m looking forward to doing this right! So happy I found this forum.

We’ll listen to that in a bit, but forum test clips should be in perfect quality WAV, not lower quality MP3. There’s no way to know if any sound damage we find is caused by the MP3, or something you’re doing wrong.

And don’t convert the MP3 to WAV. Once you have MP3 quality damage, it’s permanent.

Yes, ACX requires delivery in MP3, but if you read the instructions, it’s nose-bleed high quality, Mono, minimum. It’s a business, not sound quality decision.

All production before submission should be in WAV.

Koz

These are the DeClicker settings I suggest for that …
Declicker setting used on ''the glare of the rising sun''.png

BTW you have some distortion on that : e.g on “ammunition”, you’re a bit too close to the mic IMO.
(If you back-off a bit there will be less click-type noises too).

Here’s the WAV version (exported, not converted). I had to shorten it up to get under 2Mb. Thanks!

I think your voice sounds great! I agree with your solution to upgrade Audacity and start over. These guys did a great job reworking the mastering process. Once you use it, it could very well cut your workflow by 60 to 80%. I also noticed your RMS level was around 29dbs. Your noise floor is great! You have plenty of headroom for your 3db max. Once you upgrade and if you can keep your levels the way they are, you will simply need to adjust your normalization for your 3db max to come into ACX compliance. If you are not doing so at this time, try and keep your mic no more than 2 to 3 inches from your mouth. This will help your with your RMS levels.

While playing back or recording on Windows at least, Ctrl-M can be used to set a point-in-time (rather than a range) label. Pretty useful, so thanks for the heads up on labels! When I flub a line and have a do-over, I can just hit Ctrl-M, re-read it, and keep going. Perfect!

Ctrl-M can be used to set a point-in-time

Totally. I got the control codes mixed up.

I had to shorten it up to get under 2Mb.

Correct. No, you’re not welcome to post a novel chapter or small book.

Here’s the WAV version

You’re sure that’s your raw reading? You missed the part where you included two seconds of freeze/hold-your-breath Room Tone at the beginning of the clip. That’s not an accident. There’s a lot going on in that two seconds.

There is a brief silent snippet at the top of your clip, but it’s suspiciously perfect. Did you by any chance create that little snippet with Control-L/Create Silence? Nobody can record that work at home and most laboratories can’t do it, either.

What’s your microphone and how do you have it connected to the computer? You don’t have to use a computer to record your voice. About a quarter of my sound tests don’t originate in a computer and we’re finding a number of AudioBooks don’t either.

Koz

I did submit a sample to ACX but haven’t heard back from them in weeks

Apparently, it’s required on your eighteenth birthday to submit a test to ACX, so they have several million New Submissions to crank through. The technical standards can be handled with a computer program (the way we do it), but they also have a second pass where they test for theatrical conformance. That one is very likely a human and that’s where you get killed if you can’t read anything longer than a stop sign and your mouth hasn’t worked right since you were 9.

This is also where some computer noises get flagged. There are a couple of noises that will pass ACX Noise, but still sound like fingernails on blackboard. Such as fingernails on blackboard. Those will fail the second theatrical test.

Then there’s a test they call overprocessing. That’s when they fail you if the only way you can pass ACX is beat your voice with a stick until it bleeds. Cellphone voices need not apply.

The goal, they say explicitly, is normal, pleasant, conversation voices. That’s why our goal was to see how few corrections we needed.

Koz

I’m using a ATR2100-USB Audio-Technica mic plugged into the Ausus laptop via USB. I have the laptop just outside the walk-in closet when I record from inside. I think I remember now (recorded this a while ago) that I copied my processed room silence into the beginning of every track/chapter, so the beginning is processed. Hence, the only true ‘room noise’ is what’s between words or sentences. Sorry about that, I completely forgot I had done that back when I started.

processed room silence

I believe that.

We can’t use the spaces between words because they’re all taken up with human breathing.

That stupid 20 second voice test is harder than it looks.

I Expect that first two seconds to have some noise in the -60dB to -70dB range. What kind of noise depends on a million variables, but sound goes up and down by half and double every 6dB, so -60dB is really tiny.

Drag-select a chunk of that two seconds (avoid the extremes) and Analyze > Contrast: Measure Selection > Write down the Volume number. That should be the same noise number that ACX Check is going to give you. > Close.

Each time you use this separate test, you have to press Measure Selection again. It’s not automatic.

Koz

A side note to the side note:

they’re all taken up with human breathing.

Do Not go in and carefully excise all the breaths. ACX expects normal humans to be reading this stuff and normal humans breathe.

Given if you have verbal ticks or other annoying noises (holding up hand), then yes, you might need hand-patching or the DeEssing and DeClicking tools. Worse, Noise Reduction seems to make these traits worse.

Koz