Why is RPG Maker so dang loud? Well, the results are inconclusive.

So this was sort of inspired by @caethyril and @bgillisp unintentionally. They aren't aware of the thoughts they planted in my head.
It started when bg was showing a screenshot of Theme6 in Audacity and I was surprised to see that the wave form had quite a few peaks and valleys. For such a loud song I would have expected more of a square wave.
Then something caethyril had shared before clicked in my head. She had shared a plugin that involves a compressor built into NWJS. So it dawned on me: What if the compressor is applying some sort of excessive make-up gain after gain reduction? make-up gain is pretty common in most modern compressors, so wouldn't be surprising if they just used an existing compressor library as the base of their compressor code.
The only way to tell is to compare wave forms from the original Theme6 ogg file and the output Theme6 when launching a new project game.
Come a long friends while we walk the path to no where.

---

So my first attempt was sort of bunk. I didn't realize that my system volume mixer (at the OS level) was turned down for RM games, so the resulting recording was quite a bit quieter than the regular OGG.

However, at first I didn't consider that fact and so I just decided to try to match the ogg's volume with captured audio's volume. So I turned up the game version and turned down the ogg version until measurements matched as close as I could get on perceived volume, then rendered them out as new wavs and then compared the wav form and measured their loudness.

The results are...inconclusive.


You'll need to zoom in the picture, but you can see my loudness analyzers and the results they got. The top is the game audio and the bottom is the ogg audio.


Some of you may not understand loudness. Like you know when a sound is loud, and maybe you know we measure sounds in decibels (dB ), but the truth is that dB are sort of misleading. The subject of perceived loudness is complicated and I am not qualified to explain it, but it involves a lot of psychology, physics, and the shapes of our ears, but basically dB are not a reliable way to measure volume, which is why LUFs (Loudness units relative to full scale) became a standard measurement of perceived loudness.
So just a reminder: true loudness is not the same a perceived loudness.
Basically all you need to know is that the higher the number is for LUFs, the louder the perceived loudness is.

---

So it may be hard to get a sense of what we are talking about when it comes to "RPG Maker is loud".
So, in order to establish a baseline, as well as see what other engines/developers are targeting for their volume, I sampled 8 different games. Actually there was an interesting discovery I'll talk about as we go through these games.
Before jumping in, let me explain how this test was set up and what I am comparing.

• I would play each of these games for roughly 20 minutes while recording. I recorded everything at max volume (I double checked sliders too to ensure everything is at 100%).​
• I would then take the audio track of the recording and trim it down to the loudest section of the game (based on wave form). This section would typically be roughly 40 seconds to 1 minute in length.​
• I would then compare this sample of game play audio against Theme6. To ensure measurements averaged out properly, I would let the whole sample play through in length. The screenshots of loudness measures are at the start of the second loop of the play back, so all averages should be accounted for in the measurements.​

The question of "what makes one consider RM to be louder than others" turned out to be interesting. We all operated under the assumption that RM games tend to be the loudest, but actually, what this has shown is that there is often quite the range in volume.
As we progress through my exploration of the topic question, I think the thing to keep in mind going forward is that loudness doesn't paint the whole picture. With examples like DOOM and Warframe both being louder than RM, but afaik, there are no common complaints about the volumes of these games.

---

Because of the strange -6dB in gap in volume between the game audio and the ogg audio I became skeptical of my recording set up, so I checked all my sliders and settings, found the culprit and tried it again.



Here, we can see the green track is the captured audio from the game, while the blue track is the raw ogg file.
While there is some difference, it's hardly conclusive. We are starting to get some sort of a picture that something is happening to the game audio though. Additionally, while it might be hard to see, there are some minor difference in the peaks in the wave forms.

To be honest, I don't work with LUFs much, so I am not as familiar with this measurement system, so I decided to move it to a measurement scale I am more familiar with, even if LUFs are the better way to measure perceived volume.



I set up my limiters on each individual track. Once again, the game audio is the top, the ogg audio is the bottom.
We can see a more interesting picture getting painted here.
Do you notice the red square on the volume meters for the ogg version? This indicates that the song clips. The red light will stay on until I reset the measurements. So you may notice that the game audio doesn't have the red light. This further proves that there is in fact some form of compression happening within the engine which is helping to prevent audio clipping and thus distortion at an engine level (this won't prevent distortion that has been baked into the song, but at least the engine isn't the cause of it.)

Next, we can look at the measured volumes. Just above the volume meters you'll see numbers for the peak volume and the RMS (root mean square) volume of the song. You'll notice there is roughly a -1dB difference in volume between the two versions, once again proving that there is a compressor being applied.

I guess what this just proves is that the musicians and sound engineer working in RM are working with black magic to have a song so loud yet still retain quite a bit of dynamics...or are they...?

How about a couple more fun audio tests...

---

For another silly side quest, I wanted to compare Theme6's loudness compared to a song deliberately mixed to be loud. This is hardly scientific, but I think shows that sh/t can get a lot louder.


(Narcissistic Cannibal can be heard here)

Here I am comparing Korn's Narcissistic Cannibal's loudness to Theme6. Theme6 is on the top.
You can see that Korn is roughly 4 LUFs louder than Theme6, which is quite the difference.

You can also see that the wave form is far more square in shape than Theme6; there are less peaks and valleys in Narcissistic Cannibal when compared to Theme6. This is because the song has been deliberately mixed to be loud; and that is a whole debate sound engineers argue about which is called "The Loudness War". Evere since digital processing became ubiquitous the war began on who can make the loudest song.

Side Quest 2, Loudness Isn't a Bad Thing

Unrelated, but more people think no dynamics at all is bad (such as Korn's square wave song), but if the song is mixed for it, and the genre fits this type of mixing, I believe it can be a really fun and exciting way to mix audio. Thus you'll find a lot of my own music is quite square as well.

Here you can see my own wave forms. The top is an EDM, high intensity song and is quite square. However the bottom is a jazzy and much gentler song so I mixed it so that the dynamics can breathe.

Just tools in the toolbox.

So Korn gives us some perspective on a song that has been mixed intentionally to be loud. Well, out of curiosity, what if we take a more dynamic song that hasn't been as compressed and see what happens if we put that into the engine.
Let's take The Cinematic Orchestra's "Everyday"


I converted this song to an ogg and set it as the title menu theme for an RM project at the same volume setting as Theme6 is.


Once again, game version is on top, raw version is on bottom.
To me, I actually find this one more interesting. Although the compressor is reacting more or less the same, due to the song being far more dynamic, you can see a notable difference between the two in their peaks. Many of the peaks on the raw version (blue track) almost lick the top of the track. Meanwhile, you can see the game version doesn't quite reach the top of the track, showing the compressor in action.

An interesting thing to note is that, I thought I heard a slightly different EQ between these 2 versions. It could be my ears, it could be some sort of auditory illusion, but it seemed that the bass had been slightly rolled off in the game version, making some of the trebles ever so slightly harsher. And when I mean ever so slightly, like I debated whether it is actually there or not.
I tried to record an EQ for both tracks and then overlayed the EQs on top of each other to check for differences, but as far as my recording when, there was no perceptible difference between the 2 tracks...which is why I am second-guessing my ears.

Now, as a final piece of measurement, here I wanted to check the difference between the game audio and the ogg audio.
This is accomplished through phase inversion.
If you aren't aware what that is: if a positive wave meets a negative wave, they cancel each other out, right? The same idea applies in audio, where if 2 similar audio waves collide, they cancel each other out. This is actually how the active noise canceling in headphones works, by generating an equal but opposite wave from the sounds being heard by the microphones mounted on the headphone.

So, I inverted the phase of the game audio and played it at the exact same time as the ogg audio. The result is that any differences between the 2 versions will be heard, while anything that is exactly the same will cancel each other out.


It seems like there is quite a difference, unfortunately this data is somewhat bunk. This is due to the imprecise nature of how I had to handle this.
See, the game audio was just audio recorded directly from my desktop's output. The result is that there was a few milliseconds of silence at the start of the recording which had to be trimmed manually, then I had to manually sync the game audio as accurately as possible with the ogg audio. The further out of sync the audio is, the less precise this phase inversion method is going to be.
So, what does this tell us? Well, not much, more than likely. I think we can assume there are some minor differences between the 2 versions, but they'd likely be less perceptible than this.
As a bonus, we need to have a control group right? So just to prove to you what happens when 2 phases collide, here I have 2 copies of the Theme6 ogg file. They are both is perfect, absolute sync with each other (because they are identical files). One version of Theme6 has had its phase inverted. Here is the result:


Your speakers aren't broken. The result is that there is only silence, because you are basically taking 1 and -1 and adding them together, resulting in 0.

---

So, as discussed in the "Getting a base line" section of this article, it turns out that loudness doesn't really paint the whole picture, but in addition to loudness, mixing also plays a huge factor.
I did some investigating into the sample rate and kbps as well as other mixing techniques potentially used, but let's say not much stood out.
I will talk about some general mixing practices I dislike and one that I think contributes to the fatigue, but let's just here this more mellow mix I did:


Basically, it seems that the original mix, in general was loud in all parts of the spectrum, but I think the biggest offenders were the trebles and mids which make up the bulk of ear fatigue.
In most mixing circumstances, songs are mixed in what's called a V. The V mix is because the mids are dropped and the bass and trebles are raised. Depending on the type of song, this V can be excessive to mild, but generally speaking, every instrument has some sort of "mids" in them, and so all the instruments tend to bunch of in the mids causing excess muddiness.
With just a minor -2db reduction in the low mids, this cleared things up a lot.
I also targetted the trebles in making them less harsh by using a dynamic EQ, among other mixing techniques.

Let's look at the final result:
The wave form looks like this:


and the loudness looks like this:

(Since the remix is based on the ogg version in the game, the top graphs are loaded before all other effects, while the bottom graphs are loaded on the master bus [after all the effects])

---

Well, I wish I had some sort of interesting discovery to share with you. Actually, like I said before, Caethyril was already aware of the compressor, so it's not like that was a new discovery.
We didn't find out why the engine is perceived to be so loud either.
I found this testing interesting, but the end result is that all this testing sort of lead to no where, but maybe the information is valuable to you. Perhaps you learned something along the way. I spent the time doing all the data gathering, I figured why not share it, even if nothing new was discovered today.

I do think there is something funny going on with the audio engine of RM, but I think it would take more than just comparing wave forms.
for example, the loudness curve on the volume sliders is strange, where there is not much of a perceptible difference in volume going down from 100 to 70. 70 to 50 is only minor, and starting under 50 starts to make the biggest difference in volume.
I would need someone smart to look into what could be happening there, but that's not what this topic is about.

Sorry there is no super cool discovery at the end of this, but I hope you learned something today!

---

EDIT:
HUGE CONTENT DROP! FREE DOUBLE XP WEEKEND GOING ON NOW!
Added:

• Getting baseline; comparing other game volumes to Theme6
• Quiet songs; testing what happens to a less compressed/more dynamic song when put in the engine
• A SICKA/S R-R-R-REEEEEEEEEmmmIIXXXXXXXxXXxx *air-air-air-airhooorns*

I sort of had a new point to add to the conclusion, but that might come in a future content drop.

I think my biggest take away is I wish there was more native audio control just kinda up, front, and center in RPG Maker. That, or at least some demystification of what’s already there. I’ve always heard that it’s there, so the visualizations are still really interesting.

My biggest shock was MV Theme6 is not more square. :kaoblush:
I also love that we use all the same plugins, as if I did not attend the Nolan Boating School of Audio Engineers.

This has been a fun read and I had always wondered about RM's "loudness" as well but didn't have the patience to do testing like this.


...But I am surprised Theme 6 is not a square...that's very interesting

I love how everyone is confused that Theme6 has dynamics.

I agree with this.
I think Caethyril would need to correct me, but like I said, I believe the compressor is built into NWJS and not the RPG core files, so basically the programmers of RPG Maker likely don't do much audio-related programming and instead just relied on the defaults that are built into NWJS (which is also why I assumed maybe there is overly aggressive make-up gain that a user of NWJS would be expected to change)

So, with that being said, they likely don't know how/didn't feel confident enough to make audio functions more up front.

But that's just a theory.
An RM-devs-don't-know-audio Theory.

Boi! Time for me to dig through documentation!

I'm still digging through (which will take some time, as I have a full-time job now), but looking through the Web Audio API documentation, there are methods that allow manipulation of certain details (like panning). I wonder how the engine implements the API...

Link: https://developer.mozilla.org/en-US/docs/Web/API/BaseAudioContext

The DynamicsCompressorNode I mentioned before is a Web API: it's built into all (standard) web browsers, including NWJS (built on Chromium).

Digital audio is not my forte, but I do know that there's a lot of stuff on the average modern computer system that will attempt to "normalise" things for the end user (not just audio!). Thoughts:

1. Common audio driver enhancements include compression and/or noise gating.
2. Browsers might do their own processing, natively or via extensions.
3. I see no definition of "volume" in the WebAudio spec; maybe gain/pan are not simple dB offsets.

A question, you said RPG Maker is loud, but compare to what?
From my experience, theme6 sounds about the same when playing in MV's editor and in game.
Then I drag the ogg file into edge and it doesn't sound less loud.
Then I tried windows media player, it sounded just as loud as the browser.

Edit* Oh, forgot to mention, each time I turned the volume to 100%

Everything else one uses their computer for.

If I set my Windows system volume to a level where the testing dings are comfortable, then play anything in RPG Maker at its default volume, it's blasting painfully loud.

You may have something configured differently, but this is the common experience from anyone I've spoken to on the topic - I even made a plugin for developers to be able to reduce the default volume without having to adjust every system sound individually.

Yeah actually I have used your plugin.

But I didn't lie about the ogg sounds the same in nw.js and others. But you just need to turn the volume to 100%(the player, not your system volume).
You can easily test it by dragging the ogg into a web browser or windows media player and click the player's volume slider to adjust it.

It is just because the music is playing at their max volume, and the music itself is loud.

Sound. Compared to all of the sound in the world, it is of the top 10% of sound. My ears no longer work after listening to the Default MV Title Screen for hours straight. I can no longer fall asleep to white noise, since it no longer registers to me.

I miss my hearing.

If you've never accidentally been launched to the international space station after getting jump scared by RPG Maker audio, do you even RPG Maker?

So. The music is both arranged and mixed to be loud. @TheAM-Dol did not glean a bunch of new info, but his experiment indicates there is definitely some additional compression getting applied.

That actually reminds me, I still have portions of the default audio that I'm removing electrical noise from. If you're not careful, you make unwanted sounds louder too.

If I play my own music through RPG Maker, I can comfortably play most of them at the engine's 100%. The fact there are tracks that people are regularly turning down to 40% or lower says something.

Actually, this is a pretty good question.
While I did a comparison against Korn to show things can get louder, I never actually compared it to other games. I had the faulty assumption we all knew that RM is loud, so I didn't think about offering a starting comparison point.


So this is also a good point.
Occam's razor, it could just be that it is not engine related at all, and my tests sort of show that too. A -1dB gain reduction is not very perceptible, so I would not consider the compressor to be the main culprit of the perceived loudness.

When comparing the ogg file against the game's output, the loudness is almost the same, right?
So perhaps it's the most obvious answer: it's just the way the audio files have been mixed.

It's sort of inspired a few more tests I want to run, but won't have time to do it until Saturday. These tests are:

• Compare average loudness of a loud section of other games (engines) when compared next to RM.
• Compare what happens you take a quiet, dynamic song and put it into the engine.
• Inspect the general mixing practices of Theme6, such as EQ, dynamic range, bit rate and sample rate.

So, I guess, stick around. In a couple of days I'll have a part 2.

I think something to note is that some of the wackiest loudness actually comes from the default SEs and MEs. I kind of do suspect that most of the loudness is coming from how the assets themselves were created. By the time you layer a few things up, it’s just…exhausting. I’m still surprised by how much space Theme6 has, but the perceived loudness could be in part due to the selection of instruments as well.

I don’t notice too much funkiness with volume when I put my own audio assets in RPG Maker. Game devs being aware of how much total loudness (among other things) is stacking up, say…if they layer a song, a soundscape, and then have effects popping off is another entire topic worth talking about a loooooooot. And…kinda one of main things the Audio Clinic thread was made to accommodate feedback for~

This is why I want to inspect the EQ especially.
What is the first thing that comes to mind when I say Theme6? Likely the damn trumpets.
Trebles can quickly become fatiguing on the ears, so I wanted to check the balance.
A low bit-rate and/or low sample rate is going to further exhasperate this issue since the artifacts are most obvious in the trebles.

Depending on how it's been mixed, I might be able to remix it to be less harsh and see if that reduces the fatiguing nature of the song.

And I am singling out Theme6 because of the meme, but I think whatever processing has happened there is on most of the assets too.

Edit:
Also, lol, audio clinic which I basically used as the cool kids "share your own music" thread.
People, besides me, should make use of that thread.

Trumpets, cymbals like every three seconds, high strings, a little piccolo ditty. Like, that string patch is screaming and it’s doing what? Re-enforcing chord tones I can already hear? Once it drops out in the middle, there’s relief. @TheAM-Dol Also…perhaps I’d have a song to post in the clinic if I hadn’t gotten myself involved in some 5/4 Phrygian nightmare.

Interesting stuff. Maybe I'll do some checks on my end as well this weekend.

Broadcasting recommendations for audio: EU: -23LUFS(+/- 1) or US: -24LKFS(+/-2)

Theme6 coming in at -12LUFS:


(+10dB would feel like the volume is doubled. Yay perception.)

HUGE CONTENT DROP!
FREE DOUBLE XP WEEKEND GOING ON NOW!

I did those things I said I was going to do. The new content has been added to the article.

Added:

• Getting baseline; comparing other game volumes to Theme6
• Quiet songs; testing what happens to a less compressed/more dynamic song when put in the engine
• A SICKA/S R-R-R-REEEEEEEEEmmmIIXXXXXXXxXXxx *air-air-air-airhooorns*

Oh, and ignore the bug. I keep trying to delete these excess spoiler tags, but every time I delete and click save changes, they just...come back magically. There is content behind those excess spoiler tags. It's content that should have been wrapped into 1 spoiler tag, but for some reason got broken into 2?