VPX Setting, Adjusting and Analyzing Loudness - Plugins

browj2 wrote on 2/23/2020, 2:43 PM

Discussion on Adjusting and Checking Loudness in VPX11

The Peak Meter in the Mixer of VPX is difficult to work with for any accuracy; there is no indication of actual levels other than the blinking or steady lights. Finding peaks and knowing the peak level is difficult to impossible, and there is no indication of the integrated loudness of the entire audio.

Does anyone use any meters to monitor loudness of objects, tracks and master track? If so, which ones? See below for the ones that I have.

What loudness should one use, i.e. EBU R128 (-23 LUFS) for DVD\BR\MP4's for TV, or louder for YouTube (-13 LUFS) and Vimeo?

Other than exporting for use on a TV, exports for YouTube or video have different problems. YouTube normalizes to -13 LUFS, so -23 LUFS would be too quiet and the overall value would likely be raised unless there is clipping.

What should be the True Peak or absolute limit, -1 dB, -3 dB?

We had a detailed discussion on audio in this thread:

https://www.magix.info/us/forum/automatically-adjust-all-clips-volume-to-the-same-perceived-level--1222943/

I did a project recently that has a lot of narration and background music from various sources, plus some sound effects. The challenge was to:

Normalize to EBU R128
Adjust audio clips (narration) to have the same overall level
Using Ducking (damping) where narration was over music
Adjust the music to have the same perceived loudness
Avoid clipping
Have the loudness from one project to another about the same perceived level on the television so as to not have to make adjustments from one video to another
Measure the loudness to see the final result.

I didn't have the tools below, so I was making adjustments by eye and ear. The results were mixed.

Here are the two free tools (VST plugins) that I found for measuring and analyzing loudness:

Melda MLoudnessAnalyzer - comes with the Melda free bundle. The dimensions interface of the free version cannot be modified.
YouLean Loudness Meter 2 - free version, but there is a paid version with more information, graphing and exporting graphs to a png does not have a watermark.

These are applied using the Mixer and FX on a track.

MLoudnessAnalyzer screen shot on a music part showing that it exceeded the target:

YouLean screen shot on a music part showing that it exceeded the target:

YouLean showing mixed narration then music. The narration was close to the target, but the music went up. The overall integrated level was close to the target, but the music was noticeably louder than the narration, so I should have reduced the volume of the music object:

Below is YouLean with clipping. You can see the red marks at the top and the graph shows time code, not elapsed time. This clip was normalized to EBU R128 and the integrated level shows that it was very close at -22.9 LUFS. This was a relatively quiet clip with some loud noises. Thus, normalizing a clip like this to EBU R128 can cause problems. YouLean shows the problems and where they are.

Both of these tools are interesting and each has some advantages.

YouLean will show where the peak exceeded the target of, say, -1.0 dBFS, and you can see exactly where they are on the timeline. You can easily see which parts exceed the target in the short-term (3s moving measurement) - the red part. The graph is adjustable and you can zoom in on the timeline and slide back and forth using the key graph at the bottom. You can export the graph to a png, but it has a huge water mark. The paid version removes this and there are more features. Note: S = short-term loudness, M = Momentary, T = True Peak. Presets are limited in the free version. The plugin is for analysis only and the output cannot be adjusted.

MLoudnessAnalyzer shows LU - Loudness Units - rather than LUFS. Set the target in LUFS and it is equal to 0 LU. The graph has more things that can be graphed. You see the peaks, not just where the target is exceeded. The free version has presets for YouTube, Spotify and iTunes. There is an auto-gain that will adjust the gain to hit the target, or you can adjust it yourself. Just remember that this is for the entire track, not an object. The width of the graph is limited but has a zoom to show more or less duration. Turn off the left chart and the graph is larger, adjust the zoom level and run again.

Again, these run at track level and in the master track, but not on objects.

Another point on the measurements. The integrated LUFS measurement is gated, such that parts with very low volume are not considered in the calculations.

Damping

I did damping on the music where there was narration, reducing the level by 12 dB. This caused me some additional problems that I am still trying to figure out. I noticed that the difference in the integrated LUFS between the narration and music was not 12 dB or 12 LUFS in some cases. I think that this was because I made some adjustments to object levels after ducking. I didn't have these plugins when I edited the video, unfortunately.

Additional Tools

As noted in the other thread, there is the free Levelator, but it is not a plugin. A similar one that is not free is Auphonic. These would be useful for levelling the narration and require an export of just that part. I haven't tried either.

John CB

Comments

CubeAce wrote on 2/26/2020, 3:32 PM

@browj2

Hi John.

Beyond not wanting to introduce distortion into your audio, to what end is this important to you?

I ask for several reasons because a lot of the old standards are increasingly becoming redundant due to the current technology almost makes standards agreed upon even a short while ago are generally being ignored by those who originally thought that having standards would be a good thing.

For instance, EBU R128 is now often ignored by broadcast stations and not an enforceable standard. Other broadcasters may ask for their own guidelines to be followed such as the BBC which is mostly a law to itself.

BBC Guidelines General.

So beyond trying to make the audio intelligible at all times and not annoying by changing levels all the time what do you want to do?

Your only real restrictions are to avoid clipping and have a working dynamic range that most power amp sections find they can work within, which depending on how that is made and how old the circuitry, could be as little as 69dbA to as high as maybe 112dbA. But you are never going to get that amount of headroom in most video formats for export.

You are mainly limited though by the export file format in the end and it's encoding restrictions which to my mind are somewhat restrictive compared to what can actually be recorded, but for most instances can still be acceptable.

The main problem I have is not overly compressing or limiting material too steeply as it can suck the life out of any soundtrack.

The main thing is never to try to mix using headphones. Always use speakers. When you finish mixing try listening to the whole thing in mono on a very limited audio playback device. Something small and tinny such as a phone. See what punches through and what gets lost. With the use of limiters and their like I would always be gentle and I prefer quite slow attack vectors with short release times. There are a lot of good masterclasses for mixing audio from major studios on YouTube some of which are over three hours long but if you don't have a good audio setup at your end you may miss the points they try to make as your system struggles to show the subtle differences small adjustments can make.

As far as how close to clipping can you get? is literally that, as close as you can get because when it gets further truncated down when exported as a video file it will be even more crudely compressed and flattened if not containing enough data in the stream to start with. Uploading to sites like YouTube only seem to compress files further where there seems to be least detail in either video or audio. While there is always some degradation as transfer of such files takes place and the sites do their own thing to the uploaded files, in general if the source files are good then the returned viewable video is not that far from the uploaded version.

At the end of the day though unless you are trying to sell the end products or avoid overloading a record-able disk of some sort, then you have volume controls you can use to get your overall listening levels and I personally would not worry too much as whatever we do today will have next to no significance in the foreseeable future. You can't improve over your source files, you can only hope not to degrade them too much in transferring to new mediums.

Ray.

browj2 wrote on 2/26/2020, 7:12 PM

Hi Ray,

Didn't you read the first paragraph that I forgot to include? No? Then I had better add it.

I finished a video last week that contains a lot of narration and music from different sources. I had a lot of trouble balancing the loudness to get narration and music between narrations to be about the same perceived level. The key is "perceived" level. When watching it play on the television, I noticed that sometimes the narration was a little low, and some of the music was noticeably louder than other music pieces and the narration on either side of it. I used the built-in ducking (damping) feature for the narration when there was music. Some parts just have narration, no music. I was wondering how to get everything balanced without only listening to it, which obviously didn't work out, and started looking into a way to measure the loudness. We quickly get ear fatigue and it's hard to determine if what you heard 5 minutes ago is louder, softer or the same as what you hear now. Distortion is not a problem.

To go along with that, after watching three separate videos one after the other but created at different times (months apart), one was noticeably louder than the other, so I had to adjust the volume. One was noticeably softer than normal programming. I wanted a way to measure so that I could get all future videos output to about the same perceived loudness.

I did a fair amount of reading and watching tutorials on the subject and they all pointed to the key - getting equal perceived loudness, and several of them pointed to tools to measure loudness.

EBU R128 is just one of the standards, with integrated loudness to be -23 LUFS for broadcast. YouTube, Spotify and others have their own levels. YouTube, apparently, uses -13 LUFS. Of course, nothing should exceed 0 db. YT will decrease the gain in any videos that exceed their limit. That is ok. However, I read that they do not go the other way. Doing so could ruin the sound, but your audio will likely be lower than the next video that you or someone else watches on YT. I believe that some others will boost the sound.

There was a similar question on the Vegas Forum from someone who had loudness from different types of audio sources all over the place and was looking for a way to balance them. He did not get a good solution.

Another one on the Vegas Forum just wanted his voice recorded with the video recorder to be at the same level as the music. Normalizing both is usual solution, but it can have problems. See below.

So, how does one go about measuring the loudness to get approximately the same loudness from video to video, and even within a video for balancing voice and music? The tools described in the initial thread assist and they are not hard to use. However, loudness adjustment is definitely not everything that needs to be done, but doing so certainly helps get the levels more even.

Problem Case:

In one recording of narration, I noticed that the measured level was lower than the others, but I couldn't raise the level without clipping. I had some plosives and the first syllable of several words was far too loud. I could see this in the graphs of the tools, and simply put a volume curve on the object and reduced the volume at the peaks. Then I could adjust or renormalize and adjust the audio clip to get the same perceived level as the other parts. The tools allowed me to measure this and fine-tune. The graphs also show if a particular passage is louder or softer than others; no need to use the ears.

I'll post a graph of one of the obvious problem areas from my last video so that you can see visually that there was a problem that I did not hear during editing.

John CB

Craigster wrote on 2/26/2020, 10:05 PM

Browj2 --

You are asking some great questions.

I just did a Memorial video that included only 1 camera (so video editing was easy), but for audio included:

- primary speaker (pastor) with a lapel mic

- 4 different speakers on a condenser handheld podium mic (different heights, distances)

- Male soloist with live piano

- Female Duet with Audio Track

- Camera mic from balcony 60' away for 2 periods of 2 minutes where the CD-board recording was missed.

So, this required separate tracks to adjust track/clip gain, plus EQ and Compress each distinct sound source, and a master track limiter and room reverb to tie everything together. In this case, my limiter had necessary metering and true peak. So, I was able to do it all in VPX. In this case I set it to LUFS 14.

Many times, to be honest, I prefer to offload all audio tracks to Samplitude Pro X (I think you are also a Samplitude user, right?), where I have full disposal of all my plugins and metering and full-featured mixer. Then I just bring a master track back into VPX. It's so easy now with the audio-match features. Actually, due to distance, I generally have to manually tweak a bit, but it gets me 99.5% of the way there.

For straight-thru (i.e. non-stop) recording of multi-track audio for live band events, I have run both programs simultaneously, with audio fully in Samplitude via ReWire and Timecode. Took me awhile to get it set up, but I loved the workflow. Planning to do this more.

Best wishes.

Craig

browj2 wrote on 2/27/2020, 12:18 PM

@Craigster

Hi Craig,

Yes, I have Samplitude Pro X3 Suite. I took a look at the Loudness Meter and there is no graph, only the readout, so the two other tools would be helpful. The YouLean Meter can be adjusted to the desired level with the paid-for version.

You mention doing everything in your Memorial project in VPX and that your "limiter had necessary metering and true peak." Which limiter were you using?

You also mention that you set it to LUFS 14. How did you do this?

"For straight-thru (i.e. non-stop) recording of multi-track audio for live band events, I have run both programs simultaneously..." You mean both Samplitude and VPX? If with VPX, why?

As for sending audio out to Samplitude and then back to VPX, that is a different topic and I would like to pursue that in detail in another thread. I'll probably open the topic in a few days. Have you used VPX and Samplitude together using LoopBe1?

@CubeAce

Hi Ray,

To simplify things visually, I imported the video into SoundForge Audio Cleaning Lab (upgraded version of Magix Audio & Music Lab or Audio Cleaning Lab) and metered the audio using YouLean.

Entire track:

The sound track was mixed down and normalized to EBU R128 (-23.0 LUFS) in VPX before exporting, if I remember correctly, and the measurement shows -23.1 LUFS, which is very close. True Peak Loudness is -3.9 dB, so I could only increase the overall gain by 2.9 dB, respecting a max. of -1.0 dB true peak. To get an even higher loudness value, I would have to go back and find out where the peaks are, and lower them, or use a limiter. The problem peaks were in the narration, not the music.

Partial with Narration and Music:

Integrated level was -20.8 LUFS, so louder because the music level was set too high vs the narration. Here you can plainly see that the narration is at a lower perceived level than the music. The music is well above the line, with the area in red; the narration is almost all below the line.

Just Narration:

Integrated level was -23.7 LUFS, TP -4.7 dB. I could bring this part up by 3.7 dB if I wanted a higher level.

Integrated level was -19.2 LUFS, clearly showing that the music is much louder than the narration. Thus, I should lower the gain of the music clip by 3.8 dB.

To have equal perceived loudness, narration should be brought up slightly, music turned down.

I'll make another post to show the graph with the audio normalized to peak in SFACL.

John CB

browj2 wrote on 2/27/2020, 12:46 PM

Below shows the peaks after normalizing to full level in SFACL:

Integrated level is now -19.2 LUFS, TP = 0.0 dB. The warning level was set to -1.0 dB so I can now locate the peaks and fix them if I want to get the higher overall level.

Below shows more peaks if I change the warning to -4.0 dB:

Below shows a zoom in on the chart with a peak located by time. This one happens to be at the beginning of some music:

If I want an overall loudness level higher, then I would have get rid of the peaks before being able raising the overall gain without clipping.

To summarize:

I want narration and music to be at the same perceived level. Using the YouLean meter helps me see rather than hear.
I want an even perceived loudness level from one video to the next. The only way to tell is to use a meter like YouLean.
For posting on YouTube, I want an overall loudness of close to -13 LUFS with true peak less than -1.0 dB.
For TV, I am still undecided. I think that -23 LUFS is not loud enough, but -13 may be too loud. I'm open to suggestions.

John CB

CubeAce wrote on 2/27/2020, 5:05 PM

@browj2

Hi John.

Well the first thing I'll have to admit is I'm at least 14 years behind in the tech you are currently using so I've no experience in using something like your YouLean meter. From what I've read in your previous posts you have a great deal of varying material quality wise in one project. That in itself is going to make anything you try to do quite difficult as noise floors will vary between clips as will the frequency responses, harmonics etc that will have an effect on how they sit within the mix. I do know that if you are trying to do any form of auto ducking of a narration over a stereo music track you need to be using a compressor that has a side chain for it to work properly. I don't know if your compressors have this or not or used one or not but it can be a useful tool to keep these perceived levels in line with each other. The other thing that come to mind is tracks with lower signal to noise ratios mixed with tracks with much higher snr's may need to be gated at the lower volumes so that no signal processing occurs at the lowest volumes within that track. Without examples to listen to it is difficult to advise what to do and my lack of knowledge of more modern plugins and their abilities may lead to me not knowing of a plugin that could automate some of this work rather than having to manually adjust as you go along. I must admit I admire your dedication to trying to do all of this but you have not set yourself an easy task.

I do concur with Craig's assessment though, that if it is at all possible to export all sound tracks into an external audio editor, that would be the best way to go. From there you may even be able to set up some form of mixing template that would eventually give you more reliable repeatable exports from one project to the next.

Ray.

browj2 wrote on 4/10/2020, 12:28 PM

Just thought that I would update this as I replied to a user on the French forum about how to get fairly even perceived levels from one audio recording to another, mainly with songs.

Using MEP2019, in the image below I have two Magix audio files. The first, Snowfalls, is one of the free songs; the second, Happy, was created using Soundtrack Maker, Sportive (Big Beat). I did not normalize these, but both showed hits up to almost 0dB.

(Ignore the rectangle in the image of the meter)

Snowfalls - soft music - Integrated = -17.2 LUFS

HAPPY - (Soundtrack Maker Sportive) loud music - Integrated = -10.4 LUFS, the music sounded much louder than Snowfalls and the meter shows just that:

HAPPY - reduced volume using the middle handle to -6.8 dB, giving Integrated = -17.3 LUFS:

This made both pieces sound to be about the same loudness.

Since I started using the YouLean meter to check the audio and then adjust the volume (and other effects like compression), I have had more success in getting a more even perceived volume of both narration and music throughout a movie.

John CB

Craigster wrote on 4/15/2020, 8:22 PM

@Craigster

Hi Craig,

Yes, I have Samplitude Pro X3 Suite. I took a look at the Loudness Meter and there is no graph, only the readout, so the two other tools would be helpful. The YouLean Meter can be adjusted to the desired level with the paid-for version.

You mention doing everything in your Memorial project in VPX and that your "limiter had necessary metering and true peak." Which limiter were you using?

You also mention that you set it to LUFS 14. How did you do this?

"For straight-thru (i.e. non-stop) recording of multi-track audio for live band events, I have run both programs simultaneously..." You mean both Samplitude and VPX? If with VPX, why?

As for sending audio out to Samplitude and then back to VPX, that is a different topic and I would like to pursue that in detail in another thread. I'll probably open the topic in a few days. Have you used VPX and Samplitude together using LoopBe1?

The YouLean meter actually looks REALLY helpful. Nice find! What does the Pro version add?

I use a number of limiters for different things. But lately I have been particularly enjoying the TDR Limiter 6 GE. It includes a loudness meter. Though you have to toggle thru different options. I tend to land between -14 to -16 LUFS for my project masters. It is adjustable to taste. (I have also adjusted Samplitude's Program Loudness to -16 (instead of -23 LUFS).

The band recordings I was referring to include Video. So, I find it helpful to see video when I'm making mix decisions for a video. Highlight entrances, faux-match panning, etc. The eye can draw attention to featured parts. I have found that if I'm mixing audio to video and don't use a visual reference, that I can inadvertantly send conflicting messages. (For example, if you see on stage 3 singers, left to right: Alto, Lead, Tenor, but mix them apart from the video with alto panned partly right and tenor panned partly left. Or acoustic and electric guitars reversed, etc.).

Yes, I purchased LoopBe (actually the 32-version, though I have no need beyond the LoopBe1. Just didn't realize at the time.)

browj2 wrote on 4/16/2020, 7:45 AM

@Craigster

Hi,

The YouLean meter only measures, it's not a Limiter. I suggest that you look at their web site and this . For me, the graph is the main feature. I can see clearly parts that are too loud and also locate points where the max loudness target is exceeded on the timeline by time code.

With the paid version (Pro), you get:

Streaming presets
Dynamics graph
PDF, PNG or SVG export
Saving custom presets
Text scaling
Dark and light theme support
A/B save states
Drag & drop analysis

Or, you get:

Access to the additional presets for set the target and you can adjust it yourself;
Access to all of the various parameters that you may want to set (but probably not);
Good graph export without watermark;
Can drag and drop a file on the app to run the analysis with graph.

As I mentioned, an alternative is Melda MLoudnessAnalyzer, free but you can pay if you want, that has more flexibility in the settings. MLoudnessAnalyzer output gain can be adjusted, whereas there is no adjustment possible in YouLean.

Try both.

John CB