Overprocessing? Sure sign of overdoing it ???

Audacity Help Forum Windows
1

I am Proof-listening a new member’s spoken audio track. I do not know the member at all, nor the o/s nor if he is using Audacity, let alone the version.
Too I am too new to Audacity to know how to use forensic tools to try to see what has been done.
Untitled.png
The 8-minute track passes Librivox’s “Checker.exe” program, so it fits within the LibriVox profile, and I could understand every word, so I passed it as PLOK.
But when I was playing it back, I noticed that the waveform was not symmetrical around the centre-line of the band.

I can’t ever recall seeing a waveform that looked as if a tractor-mounted hedge-clipper had been used to trim the top!

If I had to bet money on it, I would say that the new member, perhaps over-keen to make the track sound as good as possible, used every free Effects tool to polish up the audio.

If any of our experts here, in glancing at the waveform, can suggest one or two most likely causes for the hedge-clipper effect, I would appreciate it.
I might then cautiously ask the newcomer if he has used “this Effect” to process his file.
More and more I am arriving at the opinion that, for LibriVox, there is no need to apply Effects to track unless the Proof-Listener suggests that something specific be done.
Thanks
Chris

2

https://forum.audacityteam.org/search.php?keywords=asymmetrical

3

That’s a bit “odd” (worse than what I’ve seen) but I don’t know of any common effects/processing that will do that. The person could have an unusual voice, or there might be something weird with the microphone. A high-pass filter (which is actually part of the [u]Recommended Audiobook Mastering Process[/u]) will probably make it “look better”. (A symmetrical waveform should make it “easier” to pass ACX.)

A [u]DC Offset[/u] is similar except with DC offset silence is also shifted and that’s a worse problem. (DC offset is usually caused by a hardware defect.)

4

I think you hit one of the conditions that looks wacky, but isn’t. Non-symmetrical speaking is normal. I used to be able to tell which television news anchor was speaking by which way the peaks went. Hers went down, his went up.

I could understand every word

You may need to get a little more exacting than that. You can understand every word in a telephone call, but that won’t pass theatrical testing. You wouldn’t want to listen to a book in that voice.

You should be using your high-quality, wired headphones to inspect quality, and the work, as a rule, should sound like someone telling you a fascinating story over cups of tea. It should not sound like Skype Voice, Bad MP3, Overprocessed Multi-Point Zoom, ratty internet connection, or crazy noise reduction.

One of the odd processing conditions is to Noise Reduce, Silent-Sense, or Noise Gate the voice to death and when they stop speaking, the background sound drops to dead zero. That’s not normal, and it can give odd word distortions. Up-cut sentences and clipped ends of words. Our analyzer, ACX-Check, has a warning about that.

hedge-clipper had been used to trim the top!

That’s not what you have in the illustration. The illustration is normal non-symmetrical speech where top and bottom don’t match. It is possible to have a broken microphone and get that hedge-trimmer look. Dead flat tops, not just reduced expression. You can hear that. That will sound odd, harsh, and crispy.

There’s no handy collection of tools to catch theatrical errors. What’s why you, the human (we assume) are listening to it. I have a stupid joke that you’re not a good reading candidate if your voice scares the horses. There’s no tool to detect that. You have to find a horse.


It is a little disconcerting that we’re teaching a quality control practitioner how to quality control. Wasn’t there any training or course of classes?

You can post up to 10 seconds of perfect quality WAV sound file on the forum. Post some and we’ll listen to it.

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

Koz

5

There are Quality Control tricks.

Get the performer to crunch up a newspaper in front of the microphone. Be careful not to overload the sound channel. Play it back. A broken microphone may give that hedge-trimmer effect. A regular microphone won’t.


Koz

6

Thank you Doug, for increasing my understanding.

The original recorder (“reader”) says that he is not doing anything extraordinary, but then, what is ordinary to him might be …

The recording is not for ACX (which I gather is Amazon) but if for Librivox, which has different standards and different goals, so there is no question (in my mind) as long as the spoken text is comprehensible; I have already passed the track as OK, because I can understand every word. I would not have raised this topic except that I was playing it back in Audacity and could see the representation of the wave form, a waveform unlike any I had previously seem.

In a private message to the reader I mentioned the (to me) strange waveform (at the time I did not know anything about the hardware or sofware platform he was using, recording with Audacity or something else …) so I will return to that PM and see if he wants to investigate further.

I do not (yet) fully understand DC Offset and Normalize, but here is the “MP3Conversion” macro I have been using Lo! these past three months; I did not write this macro.

Normalize:ApplyGain="1" PeakLevel="-1" RemoveDcOffset="1" StereoIndependent="0"
ExportMP3:

Thanks again,
Chris

7

Thank you, Koz. This is re-assuring. I know that when i made my first recording I felt obliged to process the life out of it; I have since learned to just speak into the microphone and let the Proof-Listener tell me what, if anything, “needs fixin’”. I found nothing audibly incomprehensible in the recording so passed it “PL-OK”.
And you are spot-on with your comment. It LOOKED wacky to me when exposed in Audacity, but then - you are well aware of my lack of expertise!(grin)

You may need to get a little more exacting than that. You can understand every word in a telephone call, but that won’t pass theatrical testing. You wouldn’t want to listen to a book in that voice.

Quite so, but Librivox has a lower set of standards of audio quality and a higher set of standards for participation “Join in and have fun recording”, and that’s Librivox.
There are some audiobooks that make me weep, but as they say “If you think you can do a better job, then re-record it and we will publish it”.

You should be using your > high-quality, wired headphones > to inspect quality,

Ah! If I but had such a set. I figured to see if I could survive a year before spending my miserly pension on state-of-the-art stuff. In six months time I will buy one of your cast-off sets (at a reduced price!) and that will be better than anything I could find in this town!

hedge-clipper had been used to trim the top!

That’s not what you have in the illustration. … It is possible to have a broken microphone and get that hedge-trimmer look.

This to my mind is worth pursuing - if the Reader is interested; it is his call. If I hadn’t glanced at the wave form i would not have raised this issue at all.
Example: I know A.A.Milne’s poems by heart (from my 75 years of being a small child), so I can PL those while sitting out in the garden, skipped verses, repeated phrases etc. Anything else I sit at the laptop and read the text along with the reader, hence I was using Audacity on this track.

… we’re teaching a quality control practitioner how to quality control. Wasn’t there any training or course of classes?

Not at all, and the PL business is hit-and-miss. If your (volunteering) PL insists on no gaps greater than two seconds, then you write a macro to achieve that:-

Comment:_="Locate gaps of two seconds or more (Greater Than 2)"
SelectAll:
LabelSounds:measurement="peak" post-offset="0" pre-offset="0" sil-dur="2" snd-dur="60" text="##1" threshold="-40" type="between"

A different PL might allow up to four seconds which, in a 474,000 word tome on logic, would let you use the length of gap to highlight the delimiters of semi-colons, sentences, paragraphs, major sections of a chapter etc.
It is the luck of the draw, and personal whim.
To make matters worse, i suspect that some PLs love to recommend Effects processing, while others do not.

This is not MoTown (grin!)

You can post up to 10 seconds of perfect quality WAV sound file on the forum. Post some and we’ll listen to it.

Thank you for this oft-repeated offer. I have never tried it, but will encourage the reader - if he is interested - to post a segment.
And in writing that I realize that I should make use of the offer; I record using the built-in microphone (don’t scream so loud, Koz) on the principle of Campbells Law on hardware faults and failures “Eighty-three percent of the time, It’s The Cable”.

Cheers, Chris