Chris’s compressor does not work quite like typical compressors.
Usually a compressor will have a “threshold” level, where audio above that level is compressed and audio below that level is not compressed.
Chris’s compressor does not do that, at least not in the conventional sense. In this effect the amount of gain is calculated taking into account the surrounding audio. A weakness of the algorithm is that it can’t work out if the end of the selection is noise or just low level audio. so it doesn’t know whether it should be amplified or not. The probably may be overcome by giving the plug-in a “hint”. Before applying the effect, if you generate a short section of high level audio (generate a tone). The plug-in will then see the high level audio which will help to avoid the rising level of the noise. The tone can then be deleted after applying the effect.
Other steps that can help to avoid the problem is to ensure that the “Floor” setting is substantially higher than the background noise level, and set “Noise gate falloff” to a small positive value. The value should not be set too high or it will perform too much gating which may produce noticeable modulation of the noise floor.
Didn’t someone post that all you really had to do was jam silence at the end of the show to give Chris something to chew on?
And yes, producing a show with pumping noise is going to be a very serious problem. I’m not sure how to deal with that. Pre-production test? Run this plugin and if the number is above XXX, the automatic ACX processing is not for you?
I sent you the mp3 via email. I think you asked for a WAV file in a previous message, but it’s 39mb so it won’t send via email. I’m not quite sure how to get it to you.
Also, you will probably cringe in horror at the content of the mp3 if you’re not a cat lover.
Thank you. I’ll look for it and cross post. We need a WAV file if we need to rip it apart and there are technical problems. I just want to see how loud you got during the voice capture. I can do that with a nice MP3.
Koz
I heard back from ACX and I posted a request for a WAV of the raw capture show or as much as will fit in the email — before any filters or effects were applied. Xina’s posting, I believe is a copy of the ACX submission. Cross-posting shortly.
Okay, finally heard back from the ACX folks. As expected, there were “issues.” The main one is:
Problem: 1. Audio does not meet ACX Mastering Requirements – Audio volume is too loud.
Solution: 1. Revise title to meet ACX Mastering Requirements - Click Here to visit our Video Lessons page which has a section on mastering. http://www.acx.com/help/acx-audio-submission-requirements/200485520#RMS
So I guess I need to know how to lower the volume to the correct level. Their specs say:
Measure between -23dB and -18dB RMS.
Why? - Put simply, this means the files fall within a specific volume range. By keeping all files within this range, listener experience is enhanced – not too loud, and not too soft. In being consistent, listeners won’t have to constantly adjust the volume of their playback device.
I know this is probably a monumentally dumb question, but when I look at the waves (see screen capture) I’m wondering if I need to make sure that the peaks fall under the 0.5 and -0.5 measurements located in a vertical line next to the waves (under the ruler point 0.0)? Also, how do I meet their specs? I ran the “Compressor” effects as a test and it does seem to lower the volume. Not sure about the settings, though - I just accepted the defaults since I don’t really know what the threshhold is, for instance).
Those replies from ACX seem to me always a little bit fuzzy.
There’s for one that they do not say how much the RMS level differs from the specification and how they do measure it in the first place.
It is absolutely no problem to forge a Analyzer that meets their standards.
Steve has posted some more details, I have to look them up.
However, if your volume is too high (in the RMS sense), compression is the opposite of what you want.
I am not sure what you mean by the 0.5 points in the display. This is probably -6 dB. Aligning with this line doesn’t help.
It should be under 0.7071, i.e. -3 dB.
We’ll do the obvious first:
Normalize the (mostly unprocessed) audio to -3 dB.
That’s the peak level they want.
We now have to measure the RMS level. As I’ve mentioned before, this does not always give the same result for all methods employed.
Steve’s Wave Stats plug-in might be a good starting point.
We have also to measure the noise floor by selecting a quasi-silent portion inbetween sentences which shouldn’t exceed -60 dB.
Where to go from here depends on many factors.
It is mainly a question of listening to the performance wheter we use a compressor, noise gate, limiter or even equalizer.
Did you have to submit the sample as mp3?
This makes a big difference, all our accurately adjusted levels could be utterly deformed.
You could sent me your sample mp3 if you don’t mind–I love cats.
Don’t go away! Long ago it was prophesied that one day someone on the forum would have “rejection” results from ACX that they were willing to share. We have been waiting for this day to come.
We have various tools for measuring audio and there are many subtly different ways to measure. What we need to do is to match up the measurement tool with their (not fully defined) measurement, then people like you will be able to use that tool and have a reasonably reliable guide to whether your recording meets their specification.
What we need now are some audio samples. Whether you prefer to post a few short clips to the forum, or send an e-mail link via PM to each of us is up to you, but what we need is some samples of your original unprocessed audio, and some samples of the audio that you sent to ACX. Your help will not only help us to help you, but to help the many other audio book authors that use Audacity.
You guys are AWESOME! I’m happy to upload some audio to help you determine what needs to be done to my files. I’ve created a Dropbox folder with two files - one is the original WAV file that does not have all the “fixes” I attempted to apply (noise reduction, compression, etc.) and it’s got WAV in the filename. The other is the MP3 I submitted to ACX of the same chapter. I made a boo-boo and didn’t change it from stereo to mono, so that was the other issue ACX noted.
Looking forward to your feedback! Also, glad to hear that there is at least one other cat lover here and that I am, apparently, the “chosen one” who will help unravel the mysteries of ACX rejection. As a person who also writes epic fantasy I feel thrilled to be chosen for … something.
I’ll leave the link up for a while but probably not indefinitely.
This is insanely valuable for trying to nail down exactly what they’re looking for and more importantly, how they’re looking for it.
As above, we need to dig through all this and make sure we’re on firm ground before publishing recommendations. I admit, some of the comments surprised me, but then, that’s why we’re doing this.
Don’t worry if you don’t understand it all yet. I’ve said multiple times I think the ACX performer dialog is much too technical for normal humans. And if you’re having trouble sleeping, I’ll explain what RMS is. I’m better than warm milk. Really.
“The peak of the waveform as compared with the effective time-duration energy density…zzzzZZZZZZZZZ”
I just listened to the WAV performance. That’s impressive. How are you doing that? That’s a desktop computer fan whir in the background, right? Microphone? Mixer? USB Interface?
First deduction (before I’ve listened to it):
The peak is 3 dB to high.
The wave statistics give a RMS level of about 2 dB less. I guess therefore that the mp3 conversion has introduced some distortion (2 dB were clipped off the top).
Thus a normalization to -5 dB to -3 dB could perhaps result in a mp3 that has the right peak and the right RMS level as well.
However, I have to listen first and control the noise floor too.
Bye the way: although I love cats, I live actually with 3 dogs–it is not always Felidae or Canes alone.
Yes, Chris produced some real problems with this presentation.
I made up a process based on your original WAV being not that far off just sitting there.
Step one, Tracks > Stereo Track to Mono.
I did do a noise removal, but I didn’t produce a Blackness of Space reading like you did. I just gently suppressed it a little and left your clothing movements and expressions alone. Then I ran Amplify and reduced the whole thing 2dB (not 3dB). That’s it.
At normal listening volume, the background noise is almost missing, volume is normal, expression is crisp and clear, and the peaks dance around -3dB.
Open for comments. I posted all the settings I used.
The down side of this process is that the original WAV is not a basket case like some of our other postings. It’s quite good and perfectly presentable right out of the original reading.
