Adding Cover Art & Meeting I-85 Compliant Needs

[theRamenNoodle]

You're right that many podcasts (and I would hypothesis that especially podcasts using Audacity) go over an hour.

I'd rather see some way for the tool to process the audio without the possibility of frustrating the user. What if it could process in 30-minute batches, separated by silence or breaks?

Otherwise, I think a warning, "Please select less than an hour of audio at a time for processing," could be informative.

Here's something else to consider. How would it handle multiple tracks? Would it process each one independently? Would it evaluate all simultaneous audio (since layering audio increases loudness)? I think most tools process a single track at a time, except for maybe Auphonic Multitrack.

[cyrano]

You're right that many podcasts (and I would hypothesis that especially podcasts using Audacity) go over an hour.

I'd rather see some way for the tool to process the audio without the possibility of frustrating the user. What if it could process in 30-minute batches, separated by silence or breaks?

That wouldn't be compliant to EBU128, as we're talking entire show here. And the "show" could be a 15 second commercial, or a 3 hour concert recording. It always needs to be evaluated in full length.

That's also why I don't understand what you are looking for. There is no magic solution there. Either you use the system as it is proposed, or it becomes a derivative that will be useless unless kept private.

Suppose everyone accepts -16 LUFS as the podcast level to go for. What happens if that podcast gets broadcast on radio?

It will be too loud and the broadcaster's system will turn it down. That might change tone, which is just what we're trying to avoid.

Itunes and other players having a leveler that seems to work at that level, is something different. It certainly doesn't work in LUFS.

I really don't see how you could extrapolate -16 from those systems, as they are just AVC and depend on the level of the other tracks in the library. In this case, the library is the show. As you don't know what's in the library, your calibration goes out the door. And the level will be very different between a library containing classical music and one containing recent metal productions.

Either you adhere to -23 or -24 LUFS and pass, or you don't. Normalizing to -16 doesn't work, unless it's for private use.

Otherwise, I think a warning, "Please select less than an hour of audio at a time for processing," could be informative.

Here's something else to consider. How would it handle multiple tracks? Would it process each one independently? Would it evaluate all simultaneous audio (since layering audio increases loudness)? I think most tools process a single track at a time, except for maybe Auphonic Multitrack.

Multi track shouldn't be to big of a problem. Are there any podcasters streaming anything other than mono or stereo?

I certainly haven't ever found one. Not that I've looked that hard

[kozikowski]

Are there any podcasters streaming anything other than mono or stereo?

They risk becomming isolated celebrities. Everybody is listening on their phones or portable private music devices plugged into earphones/earbuds. I've seen headphones that claim surround.

I can extend the problem. In order to get an audience, you need to create a stereo-mixdown in addition to the surround. So the people listening on the massive home sound systems will use surround and everybody else listens in stereo. That's an extra production step. Producers tend to have a lot of extra time available, right?

I don't think I've used the surround service on a DVD in..... When was Bush-41?

Koz

[kozikowski]

I can think of one scenario. You're listening in VR. As you turn your head, the sound field changes.

Koz

[steve]

What if it could process in 30-minute batches, separated by silence or breaks?
Otherwise, I think a warning, "Please select less than an hour of audio at a time for processing," could be informative.

"Ideally" the way that Audacity interacts with Nyquist needs to be changed so that we can finally overcome the issue with long tracks. Realistically, that is not likely to happen in the near future, so we're looking at what can be done using the current implementation of Nyquist in Audacity.

We could certainly give a warning and bail if the selection is over an hour.

"Analyzing" and "Normalizing" are very different. To normalize we need two passes - first to analyze the audio to calculate the current level, then a second pass to apply the required gain (and optional peak limiting).
I think it 'should' be possible to analyze the level (in LUFS) for a selection of several hours.
I don't think that it will be possible to 'normalize' a selection of many hours duration as one application of one effect.

How would it handle multiple tracks?

Nyquist processes one "track" (mono or stereo).
For multiple tracks we could either:
* temporarily store the measurement of each track and analyze the tracks as one multi-channel sound. The plug-in could not "go back" and normalize the tracks without manually applying the effect (or a complimentary effect) as a separate job.
or
* simply process each one independently.

As there appears to be little call for multi-channel podcasts, the second approach looks more useful.

That wouldn't be compliant to EBU128, as we're talking entire show here. And the "show" could be a 15 second commercial, or a 3 hour concert recording. It always needs to be evaluated in full length.

To be clear, we are not suggesting that podcasts do or should meet the EBU recommendations for broadcast transmissions. We are now talking about being able to normalize "loudness" to a level specified by the user (measured in LUFS).

Regarding EBU R-128:
If we have a 3 hour show, if one hour fails to meet EBU R-128 recommendations, then the 3 hour show fails, but if all 3 one hour segments pass, then doesn't that mean that the full 3 hour show passes?

I can think of one scenario. You're listening in VR. As you turn your head, the sound field changes.

That would be cool

[cyrano]

I can think of one scenario. You're listening in VR. As you turn your head, the sound field changes.

Koz

That's what several big brands are trying to launch this year.

That wouldn't be compliant to EBU128, as we're talking entire show here. And the "show" could be a 15 second commercial, or a 3 hour concert recording. It always needs to be evaluated in full length.

To be clear, we are not suggesting that podcasts do or should meet the EBU recommendations for broadcast transmissions. We are now talking about being able to normalize "loudness" to a level specified by the user (measured in LUFS).

Why not adhere to the proposed standard?

I mean, it's not of any interest to me, personally, it just seems a good idea. Taking itunes' "noramlization" as a reference seems just wrong.

Regarding EBU R-128:
If we have a 3 hour show, if one hour fails to meet EBU R-128 recommendations, then the 3 hour show fails, but if all 3 one hour segments pass, then doesn't that mean that the full 3 hour show passes?

I suppose the show could be cut in smaller parts for calculations. I think measurement devices do it that way, in view of their very limited hardware capabilities. But I have no idea how the different calculations are pieced together. It can't be a simple average, I think.

[theRamenNoodle]

Here's my theory.

If you have three hours of audio, but you can normalize only an hour at a time, it would seem that the amount of audio within an audio would average out enough so that each of the three separate hours would be within a very short variance from each other. Note, that's after compression, which reduces the loudness range.

Sure, you might have a loud yell in one hour, but wouldn't that balance out with the rest of the hour of audio?

[cyrano]

Sure, you might have a loud yell in one hour, but wouldn't that balance out with the rest of the hour of audio?

In this example, it wouldn't matter. But what if you have an interview in the first hour and a loud rock concert in the second hour?

[steve]

But what if you have an interview in the first hour and a loud rock concert in the second hour?

So long as the loudness is within the desired range for both on hour segments, I don't think there is a problem. The standard(s) set a range for the average loudness, and a limit on the short term loudness and maximum (true) peak level. If each hour falls within the range and no part of either hour exceed the short term loudness or peak level, then I think the two hour show is bound to comply.

According to ITU-R BS.1770-4, the loudness (in LUFS) is basically the mean square of pre-filtered (K-weighted) audio. (It's a little more complex in the detail, and more complex for multi-channel sound).

Why not adhere to the proposed standard?

For what? For radio broadcast? Yes, for radio broadcast you should adhere to the R-128 recommendations.

For podcasts, broadcast standards are commonly said to be too low for the target audience (listening on portable devices in noisy environments).

[cyrano]

For podcasts, broadcast standards are commonly said to be too low for the target audience (listening on portable devices in noisy environments).

That's the point, Steve. "Commonly said"...

AFAIK nobody really made a study of it. And since we know that people are inclined to find louder always better, it's a dangerous conclusion.

So what I'd like to see, is WHO came to this conclusion on WHICH basis. The only argument so far, is that AVC systems in players seem do use that level. I only use one of these regularly, and that's the one in ITunes. And I know this one doesn't put out a specific level of loudness, but is meant to level all your songs to the same level. It's very different for a classical library compared to a rock library.

I'd like to do some testing, but it's not easy. -16 LUFS for podcasts still sounds like a folk wisdom to me.