Meeting ACX Requirements w/ Audacity

Special Interest Groups Audiobook Production
81

Here’s the thing,

Robert did do an amazing job with the clip.

But I still have no idea how to achieve that on my own.

Links to the filters I would need, as well as pictures of what switch to move where would be helpful, so that I can master something on my own, and not have to turn to the forum with the same question again :wink:

I’m not entirely sure myself about the “pops” or “clicks” when I listen to it on my end both raw and edited, the only thing I can think of if either “mouth clicks” or when certain levels pop too loud past the pop filter.

:neutral_face:

82

Here’s an example:

Yes, Ian’s work has those clicks too and that’s why isuspect the snow ball mics (or their drivers).
I hope you can hear those clicks too or I’m going to jumb in circles again (like last time with Ian’s Tin-can sound).

The 90 Hz bumps are mainly due to the pops.
I’m just trying to design an automatic pop filter for Maria, however, it doesn’t seem to work very well yet.

83

Let me ask something oblique. What tools and how were they used to determine the number of times Steve’s Limiter was applied, and why didn’t you just crank the severity of the limiting up and apply it once? Or am I missing something important?

Koz

84

I didn’t want the soft knee go down into the main signal region.
Limiting to -9 dB would have a long transition possibly down to -18 dB, whereas 3 times -3 dB does only influence the first 12 dB of the original.
The first method pushes the Rms further down and thus you’ll effectively have to limit more than you actually wanted and the noise floor comes nearer and nearer…
It’s like taking off chips from a stool’s legs in order to make it stable. You end up with sitting on the floor if you’re unlucky.
Limiting by 3 to 5 dB is probably the best compromise between energy preservation and distortion.
However, I haven’t fully tested and perhaps I’m only talking rubbish, it’s Steve’s plug-in after all.

85

To work.

Have you ever added a tool or effect to your Audacity? Have you ever added a pre-baked filter to Effect > Equalization?

The ACX process (so far) involves a number of little home-made programs and it’s something of an Easter-egg hunt to find them all and install them. I’ve been keeping notes.

Koz

86

Limiting by 3 to 5 dB is probably the best compromise between energy preservation and distortion.

I know this immediately gives everyone a bad taste, but how different is this from Effect > Leveler? I was able to get remarkably close to the required density of sound with Effect > Leveler: Moderate. I didn’t hear any great leap in harshness or distortion.

Correct me, but all Leveler does is apply a log transfer curve to the work so the gain falls away as the waveform increases. As you increase the desired effect through Light, Moderate, Severe, etc, the curve becomes steeper.

Multiple applications of Severe will produce the aircraft controller voice which is one of the ultimate jobs for a voice that doesn’t ever change volume, but you do need some “screechy” filtering to get there.

Koz

87

A little comparison shows the timbre-modifying nature of the leveller effect.
The sweep goes from 220 Hz to 2200 Hz, from -2 dB to silence.
All corrected sweeps have an Rms value of -7 dB, the original -9.8 dB.
The order is:

  • original
  • leveller (2 x moderate)
  • Steve’s Limiter
  • My downwards compressor/leveller

    The difference between leveller and limiter is that the former changes every single sample value whereas the limiter works more globally, i.e. a whole bunch of samples are attenuated at once.
    My leveller works by lowering sequences with high Rms or peak according to their length. This is especially good for voices but not for e.g. organ pads and notes with long tails.

The leveller can sometimes support a weak voice by introducing new harmonics, however, I would only use it on a separate track and mix it with the original.

88

You seem to have missed the first part of that question.

What tools and how were they used to determine the number of times Steve’s Limiter was applied…

Koz

89

Right then.

Part of the application of Steve’s Limiter is following it with Effect > Amplify to -3? Or is that only if I’m preparing client delivery?

I accepted the latest submission and ran it through what I perceive to be the accepted process, and I got the -60 noise and the -3 peaks in good order, but I didn’t get the high RMS value. It’s still-25dB or worse. Are we applying multiple 3dB Steve’s filter one after the other? Where do I set the Release Time?

Isn’t it the case that the work has to achieve at least -23dB RMS after the peak adjustment to -3dB? Do I perform the Contrast tool measurement and then extrapolate where it’s going to end up?

Koz

90

Right then.
Part of the application of Steve’s Limiter is following it with Effect > Amplify to -3? Or is that only if I’m preparing client delivery?

I think that there are two versions of the same limiter effect. Mine has just the option to bring the level up to 0 dB after completion.
Thus, the multiple applications of the limiter aim for an Rms over -20 dB (ideally -17 dB), followed by a -3 dB amplification eventually.
However, it needs a test mp3 export to determine the actual track gain. I had to set the peak level to -2.7 dB in order to get a mp3 with -3 dB after re-import and decoding.
That’s actually the part that annoys me a bit. Firstly, that Acx works with mp3 as base for further encoding and secondly that this format changes the properly set levels so randomly.

I accepted the latest submission and ran it through what I perceive to be the accepted process, and I got the -60 noise and the -3 peaks in good order, but I didn’t get the high RMS value. It’s still-25dB or worse. Are we applying multiple 3dB Steve’s filter one after the other? Where do I set the Release Time?

Do you refer to the “hold” slider? I can’t recall right away how the three settings (rise time, fall time and look ahead) are calculated from this single setting. The release time is normally 4 times as long as the attack and the look ahead is the max value of the two.

Isn’t it the case that the work has to achieve at least -23dB RMS > after > the peak adjustment to -3dB? Do I perform the Contrast tool measurement and then extrapolate where it’s going to end up?

Koz

Certainly, the -3 dB has to be taken into account.
As I said before, I run my own kind of compression first over it to bring down loud passages, e.g. direct speech and exclamations.
Although it works with Rms levels, some coinciding peaks are sometimes attenuated too, e.g. the audio goes down from 0 to -3.8 dB.
The limiter is then multiple times applied to bring down extraneous peaks. Still, the goal is -15 to -20 dB Rms (from 0 dB peak level seen).
Although the noise removal should be applied early, I usually put it off because such huge Rms changes (10 dB and more) raise the noise level again into audible regions (since the limiter has no noise floor fall-off).
Thus, one has either to reduce the noise level in the beginning with the offset added, or to do the noise reduction twice or to do it in the end, whatever fits best. It surely depends on other effects that are applied inbetween–highpassfiltering, click removal, de-essing, equalization…

Koz, I think you should build up the audiobook checklist as a flow diagram with several questions and the underlying decision branches.

91

Koz, I think you should build up the audiobook checklist as a flow diagram with several questions and the underlying decision branches.

I think the goal is a lot simpler.
Make my audiobook submission compliant [Y/N]

I’m looking at all the custom filters and effects I’m collecting and it qualifies as a career move. I have an award-winning bookmark collection between two computers.

"OK, so where was LF_rolloff_for_speech.xml again…

I’m considering hosting some of this stuff myself to get it all in one place.

Koz

92

@The_zatanna
Where are you in the process? Have you submitted anything yet? I got your latest to pass what I perceive to be AudioBook Compliance and I need to start telling you how I did it.

Koz

93

Hey Koz,

I honestly haven’t submitted anything yet. Apparently I shan’t continue reading until I submit the first 15 to see if the author approves. I’m excited to see how the steps work out into how I can achieve compliance myself :slight_smile:

Cheers!

94

Cool. Step one. Have you ever downloaded and installed an effect or filter into Audacity? Much of what we’re doing does not come with the factory Audacity program.

Koz

95

Listening on my 2010 Mac Mini with OS volume slider set at mid point, Sony MDR v6 headphones, I can’t hear the tiny clicks Robert points out @ 2:10 (15take1again.wav). I can hear them maxing out the volume slider but it’s not loud enough to distract or irritate.

Was a bit surprised to find how much my box fan turned on high, 7 feet behind me was keeping me from hearing slight echo of the OP’s first room level and mic position tests earlier in the thread. Once I turned that off I could hear a lot more detail even what sounds like low frequency fan noise rumble in the Room Tone portion. Maxing out my OS volume slider now makes the ACX headroom RMS specs make sense as a safe parameter for setting overall loudness in order to accommodate a wide range of listening devices with variable default volume levels.

Both mid/max volume settings on my system allowed both comfort and audibility.

Overall I think either one, 15take1again.wav or Robert’s 15take2again.mp3 sounds good for comfortable extended listening even with my box fan on.

96

That’s the raw version and the clicks are more prominent after raising the Rms by 10 dB via limitation.
Take the short sample file with some isolated clicks earlier in this thread.
Use the change speed effect (-50 %) twice.
You’ll here 2 clicks (or rather knocks now) before “I”, 3 after that and a nasty one at “and”.
My speakers pardon nothing of that kind.
Can you hear the pops on “pulled the comforter…” and “appreciate”?
It is certainly true that the raw version itself is quite good. It lacks some polish though.

97

The original recordings do get remarkably close to the mark just sitting there. It gives us a false sense of accomplishment when we buff it up a little and out the door. The problem (besides making it so normal humans can manage this production) is making it so somebody who created trash can do it, or alternately, know that the show failed. If you go back and forth between all the parameters and you can’t hit all of them at the same time, then you have no show.

Sometimes that’s good to know. I once was a party to a very, very serious news event and parts of my equipment failed. I radio’d to the director and far from making a fuss, they instantly worked around it.

Some presenters may need to work around it.

Koz

98

Robert, hope you can pardon my lack of patience for not wanting to hunt for the spot within the entire 15 minute presentation you’re referring concerning the pops on “pulled the comforter…” section. It took me a while to find the spot in your previous amplified pops/clicks cropped sample you posted, so it would help if you noted the time position within Audacity where these noises occur.

In the mean time I’ll take your word for it that you hear it. Whether those tiny pops/clicks are within ACX requirements of not being extraneous, distracting and unwanted noises is anyone’s guess. It would be more helpful to find out what’s causing it. In fact I hear tiny crackly sounds in your synthetic 200-2200Hz wave signal demo comparison between Leveler & Limiter. Thanks for that, BTW. Quite helpful. Other than that what do you think could be the cause of these tiny pops/clicks? Audacity encoding? Sound card/mic A/D and/or D/A conversion quantization or some kind of electrical noise introduced throughout the data capture/process pipeline?

Don’t know how many folks listen to audiobooks through their home speakers and whether those noises could be heard, probably on ear buds. On my headphones they seem to blend in as part of the speakers voice or movement of clothes, jewelery even at max Mac Mini volume. I know for sure I won’t hear those noises on my vintage '72 Sansui 2000A amp/'92 Technics CD player/'85 Norman Labs Model 82 bookshelf speakers and I certainly won’t hear them in my car.

99

Sorry I’m late, I have downloaded and installed effects on my old desktop, so I am familiar with installing new tools if that is where it’s going… :slight_smile:

100

I didn’t say that you have to hunt for the clicks yourself…
I was referring to a 2 s sample file that has the clicks from just one phrase. I’ve provided the time index for an earlier version of “15” that hadn’t the long room tone sample in the beginning.
Forget about the pops, they won’t be noticeable via head phones, I reckon.
I still think that the snow ball mic produces the clicks, if only due to the installed drivers.
However, it could as well come from noisy HD accesses that bleed into the recording.
My internal sound card’s front inputs are for instance not useable for just this reason.