I went to make a demo in my home studio today, and I thought I’d check the noise floor first. I hit record without doing a take, just to let the microphone pick up the ambient sounds. I passed the ACX check noise floor test with flying colors.
I went to record a performance take, thinking that my settings were good, but when I plugged it into ACX check my RMS was too low but the noise floor was suddenly too high? I tried again and again with the same result.
My understanding is that noise floor refers to the background noise present in a take, and RMS refers to the db of the performance itself. If that’s the case, why would the performance take cause the noise floor to spike? I thought the values were unrelated?
I’m sure there’s a simple explanation, and I would definitely appreciate the help. Apologies if this question has already been answered before. I did some digging on the forum but couldn’t find anything.
Amplifying so the RMS is higher inevitably increases the noise-floor by the same amount.
An expander plug-in (like couture) can selectively squash down the noise-floor when you’re not speaking,
but IMO only about -6db reduction is possible before it sounds conspicuously processed.
Another issue that newbies have (not you, of course) is that they think they are recording into their new expensive microphone, but later find out that they were recording into the computers built-in microphone instead. Do the scratch test. Do a Transport > Rescan and check your device selection bar:
they think they are recording into their new expensive microphone, but later find out that they were recording into the computers built-in microphone instead.
You can do a physical reality check. Scratch each one. You may need to consult your instructions to find out where your computer’s microphone is.
There is a cousin problem as well. Some computers automatically apply noise reduction or noise gating. That can give you noise that’s only there when you speak.
It’s not unusual for new users to produce a timeline that’s almost a straight, flat, blue line instead of a voice track. That can give a performance that will either pass RMS (loudness) or Noise, but not both.
That’s the long scholarly version. This is the short version.
There is a caution that mastering will only get you past the technical standards. You still have to be able to read in a pleasant and engaging manner and not scare the horses when you talk.
Do you publish stories with hero, heroine, setting, and plot? ACX likes those.
I didn’t say so above, but you can still fail noise even if you do everything right. Home readers never pass noise. -60dB Noise (the ACX standard) means your background noise has to be 1000 times quieter than your voice. My house is pretty quiet, but I still have to hide the ticking wall clock and wait for the Metrobus and construction trucks to go by.
So I managed to find a workaround with the Noisefloor. I started recording into Ableton live, a program I’m much more familiar with, and I used the built in noise gate they have been passing the Noisefloor test ever since.
Now the only problem is RMS and peaking. I’m in a constant push and pull between having the RMS be too low or the clip peak. I turn down the gain, not enough RMS value, turn up the gain, peaking. Sometimes the RMS is too low AND the clip is peaking. Baffling.
I find I can keep the values in the green zone as long as I keep my voice volume and tone exactly the same, but that makes for a less nuanced performance.
There has to be a program that can fix it right? I would think that an audio engineer created something that allows you to sink your voice low and raise it but the software would have parameters you could set so your voice is never too soft or too loud. The program would compensate. Does this sorcery exist?
And I’m a voice actor to answer your question! Ironic since I’m struggling with voice recordings, but I usually record in studio through my agency where there’s an engineer. COVID19 makes that impossible these days though.
Now the only problem is RMS and peaking. I’m in a constant push and pull between having the RMS be too low or the clip peak. I turn down the gain, not enough RMS value, turn up the gain, peaking. Sometimes the RMS is too low AND the clip is peaking. Baffling.
That’s perfectly normal before you RMS Normalize and Limit. See Koz’ post above.
So I managed to find a workaround with the Noisefloor. I started recording into Ableton live, a program I’m much more familiar with, and I used the built in noise gate they have been passing the Noisefloor test ever since.
Don’t be too aggressive with the noise gate. If you have pure digital silence (-infinity db) ACX will probably reject you for “over processing”. FYI - Audacity also has a noise gate.
It is suggested, strongly, that you apply all three tools in the mastering suite before you do anything else. Measure your results and then maybe apply touch ups here and there as needed. There’s nothing quite like having voice processing tools fighting each other.
We can give you a really good idea of the processes and techniques if you record and post a 10 second sound test.
I passed the ACX check noise floor test with flying colors.
Are you getting set up for audiobook reading? ACX has provision for a voice actor reading someone’s published book or a voice actor reading their own published book. What they don’t have is provision for reading an unpublished book. You should be reading a book we can buy on Amazon and it has to be in an acceptable group.
No reading, for example, dictionaries or cook books.
I find I can keep the values in the green zone as long as I keep my voice volume and tone exactly the same, but that makes for a less nuanced performance.
The idea behind the goal illustration is to avoid two common mistakes: So loud the sound distorts…
And so quiet that the mastering tools stop working.
Somewhere in the middle is good. We picked a peak value of about half way as a fuzzy, approximate goal. If you get to the end of your performance and your timeline looks like either of the above, it’s going to be a long day.
If you are a commercial announcer, you may be accustomed to your work on broadcast which has transmitter volume processing and corrections. Audiobooks don’t have that. You are effectively mainlining your voice into someone’s ear, so lots of theatrical expression may not be the best thing, but that’s between you and ACX. After you pass technical testing and acceptance, you will need to pass Human Quality Control where a real person listens to the work for theatrical errors.
Audacity Audiobook Mastering is not the only process published. There is one that uses multiple passes of compression to achieve acceptance. It’s up to you which you want to use. But this one, when followed, has a pretty good track record of getting people published.
Here’s a test I just made. Noisefloor and peaking pass the test but RMS is too low. Is it as simple as turning up the gain on my interface and adding the audacity noise gate?
Here’s another test I did. I ran it through all the Audiobook Mastering specifications. It technically passes the ACX check, but I feel like it doesn’t work as a recording right? Like you can still hear that microphone hiss in the silence before I start talking.
Like you can still hear that microphone hiss in the silence before I start talking.
You can add simple noise reduction on top of that, but it’s not the best idea to go “diving for noise.” If you have to crank up the volume to hear the hiss, it doesn’t count unless you want it to.
Set the volume for comfortable listening to the performance and then roll it back to the beginning and listen without touching anything.
The ACX limit is -60dB, we recommend -65dB and your mastered clip is -68dB.
Here it is with gentle noise reduction. This reduction version is handy because ACX can’t hear it working.
And through all that, the work sounds exactly like you.
If you want to go to the extra Noise Reduction step…
Drag-select a little of the “clean” room noise/background hiss (no voice or other noises). Effect > Noise Reduction > Profile.
Select the whole chapter or performance. Select button on the left. Effect > Noise Reduction > 6, 6, 6 > OK. That’s Noise Reduction of the Beast for obvious reasons.
Done. We should remember you’ll have to do this little dance after each mastering pass after each chapter. Or you can go with the excellent mastered quality you already have which is generously better than ACX is expecting.
It’s been our experience that if you can make it past ACX, you can submit just about anywhere else. Those numbers are a close cousin to broadcast proof of performance.
How did you do that? Home readers never pass noise and home style microphones come with built-in shortcomings and odd sound distortions. You don’t seem to have any of those problems.
Emailed you this info but thought I’d post it here too. If you were asking about the first take I did, my microphone is an LA320, and the pop filter was about 3 or 4 inches from the mic grill. I ran that into my interface, a Scarlet 2i2. I think I had the gain low enough on my interface to pass the noise floor check. I also spoke with my lips nearly touching the pop filter. I was sitting in a little acoustic fort made out of mattresses and blankets, which helps keep the space acoustically dry.
Wondering if maybe it’s the quality of the microphone? It’s a higher end condenser, so it can be good at filtering out unwanted noise all by itself.
The second take I did the exact same thing, but I just ran it through all the plugins in post.
