Personal blog of Elijs 'X2Eliah' Dima

Music quality in videogames: the investigation

Foreword & Brief TL:DR

This is something I have wanted to do, as a personal curiosity-satisfying thing, for a while. For years I have held the opinion that the music we get to hear in videogames is technically outdated, low-quality, glitchy and otherwise messed up – unable to meet modern standards of audio quality (meaning, hi-fi records in lossless formats with minimal clipping). But I’ve never really taken the time to check any of that. Until now. I’ve just (as of this writing) finished going through a bunch of relatively modern games, checking what audio file formats they use, with what bitrates, how their audio visualizations look in Audacity.

To save you the time, let me place this warning upfront: I am not a sound engineer. I am not an audio codec wizard. I am not even, when it comes to it, a strict audiophile. Well – maybe a bit. This is not a scientific investigation spanning hundreds of sample cases and so on. This is just a short thing I did, for myself. This post will have 15 waveform images, all in all; that should set your expectations accordingly.

Also – If you are looking for instructions how to get music out of game files when it’s somehow packed, then this is the wrong place for you. I will mention any difficulties and observations, but I won’t link the tools used nor tell you what to do in that area.

TL:DR part – If you just want to know if videogames really have worse audio quality than proper music: Mostly, yes. Unfortunately. But it is not as bad as it would seem, it is not as some of the terrible stuff we see in modern popular music, and there are definite signs of improvement over time. Perhaps five, ten years down the line we’ll have lossless / high-quality videogame audio all over the place. Now, though, high-quality music is still a rarity.

If you haven’t been scared off yet / are still interested to read more, then let’s get down to business.

The idea and the immediate problems

It would be a lie to say that getting a decent overview of music in modern games is easy. For one – there’s no way that I’d re-download every single game I have on steam/etc. I just randomly picked out the ones that I had on the hard drive, and tried to get to the music they have.

A lot of the games I checked had their music packed and encrypted – effectively inaccessible to a normal user. That is a complete cock-up, imo. If a gamer has paid for the game, and has all the digital bits needed to make sound in that game, then those bits should be playable from outside of the game, too. Some games remedy that with an included OST. Many more don’t do that, and demand the player to purchase a separate OST disc with remastered choice picks of the game’s audio files. I plan to tackle a few OSTs sometime later as well, by the way. Anyway. There was a choice – ignore all encrypted music data, or try and decrypt it, keeping in mind that any format conversion risks a notable decrease in measurement accuracy. Well.. Given that the majority of games did have their sounds encrypted, I chose to check them out as well. I will put a note saying what results are “native” and what are after an arduous process of decrypting/extracting/converting. For what it is worth, I will also, in those cases, add a subjective opinion on whether the resulting soundfiles sounded similar to in-game reproduction or not. If you don’t consider that to be accurate enough – I basically agree. But it’s either that or nothing at all. I’ll do my best to present what I found, and you can draw your own conclusions as much as you want. If you know how to & can do an audio quality evaluation on videogames properly – hell yes, please do that.

For this post, we’ll have 12 games to talk of – 10 relatively new, 2 relatively old. In addition, I’ll dump four reference waveform pictures of other, “regular” music to give the videogame-data some context. For each game, I’ll show the graphs of only one song – obviously I didn’t check them all, but I did look at a bunch to make sure that the selected sample is not an outlier.


Game the first: Botanicula

This one.. This is a bit of a disappointment, really. The audio files are accessible easily enough, to be sure – though it is quite a mess in terms of naming and layouts. Anyway. The files are in .mp3 format, with a 128kbps bitrate. That’s bad. Properly bad. On the other hand, looking at the graph below, there is no clipping or over-large use of compression, and the frequency range, given the limitations of the bitrate, is pretty good. Subjectively, the files sound okay-ish. Not an impressive depth or clarity by any means, but passable. Still – this is something that only works in the game. Out-of-game, without the distracting visual accompaniment, it’s just not good enough.

Wave Graph and Frequency Spectrum for Botanicula


Game the second: Deus Ex: Human Revolution

Okay, I hate this game now. Seriously. They have the largest extent of stupid data encryption I’ve seen. The audio files – natively in .fsb format – are multiplexed in pairs, and packed into massive encapsulation files. To get the graphs, I converted the .fsb files into .wav, so there shouldn’t be too great a loss in detail, if any. I hope. Anyway, what we see below is not encouraging. For one, the full frequency spectrum is severely underused – at fsb and wav formats, the entire upper end from ~16000Hz onwards should not be absent. Luckily, there is, at least, no clipping – but the amplitude is, again, less than it could be. Perhaps this is an artefact of conversion from an fmod sound bank to wave file. Maybe. As it is, though, it’s not suitable for enjoying out-of-game in this state.

Wave graph and frequency spectrum of Deus Ex: Human Revolution


Game the third: Skyrim

Slightly messy here. The audio files are packed in a proprietary archive (.bsa), and themselves are encoded in .xwm form (which is xWMA, which is a subset of the common wma). Converted to .wav. I really don’t know if something went horribly wrong during the conversion, or if the sound quality is truly that lack-luster. Those red lines – that’s clipping. As you can see, there’s a ton of it. And the frequency spectrum, dropping off at 13kHz.. Really? Bloody hell. Subjective note time: In general, at least on my computer Skyrim has a weird sound-signature. It almost feels as if it isn’t using the soundcard properly – the music is way more muted than on any other game, with far, far less impact. Perversely, even at these frankly horrible results, the exported .wav sounds better in a media player than the original in-game thing.

Wave graph and frequency spectrum of TESV: Skyrim


Game the fourth: Oblivion

This one is slightly older, yes. Okay, it is a really old game by now. But, I had it on my hard drive, it had it’s music freely accessible in its data directories, in .mp3 at 192kbps. Bad by modern standards, indicative of the times it was developed in, I suppose. And yet, despite all the disadvantages, you see that there’s almost no clipping – bar one glitch, and the frequency spectrum is actually wider (less saturated though). Also, more quiet. Just matching up oblivion and Skyrim seems to reflect the average evolution of music in general, really. Newer = louder, glitchier, more ‘full’. Interesting.

Wave graph and frequency spectrum of TESIV: Oblivion


Game the fifth: Eufloria

Another indie game. The audio files were easily accessible, but in .ogg format. I can’t really say at which quality rates they were created. Many of the files were actually encoded in mono, and represented environmental/entity sounds. There were very few proper “music” tracks, and of those that were there.. well, see the waveform for yourself – there’s a crazy lack of amplitude.

Wave graph and frequency spectrum of Eufloria


Game the sixth: Portal 2

Well, luckily the music was, at least, unencrypted. It was packed, of course, because god forbid we try to play the game’s soundtrack ourselves, right? To be frank, I expected better from Valve on that front. On the other hand, the audio is in pure .wav, which should, in theory, allow for superb audio quality. In reality, it is.. close, but not quite there yet. As you can see by the red bits, there’s still some clipping. That aside, though, the amplitude is great, the sound is not overcompressed, and the frequency spectrum use is – well, you can see for yourself. It’s all there, right up to 22.4kHz. All in all, I’m pretty satisfied with this, and would definitely say that Portal 2 is a game with properly good audio.

Wave Graph and frequency spectrum of music in Portal2


Game the seventh: Saint’s Row: The Third

Right, this one is terrible. The audio files are really packed in arcane structures and encrypted in proprietary forms. It is possible to extract them – and I did so – but the resulting files are ogg vorbis audiofiles, which, as you know, are lossy. So I cannot guarantee that what came out was the same as what was in game. On a subjective level, I can say that in-game the music from radio stations really did sound quite crap-tastic, so perhaps the results in the graph below are not far from the truth. As for the results.. Well, see for yourself. Almost the entire track (5th song on radio genX) is clipping, and the compression is utterly bonkers. It sounds terrible (as you’d expect). In this case I truly think that the whole process of decrypting, extracting, converting from game’s data archives has affected the result. Otherwise, the person responsible for SR3’s music should be fired asap. Perhaps, this entry is quite probably not applicable – but I’m still including the image so you see how a failed/ruined conversion looks like. On the other hand – this looks very similar to a later example of modern “regular music”, so…

Wave Graph and frequency spectrum of music in Saint's Row The Third


Game the eight: X3: Terran Conflict

This game is also a mess, but in a different way. The audio files are easily accessible – game directory/soundtrack, and it’s all in easy-to-use .mp3 files. The problem is, this game has a lot of reused music from previous iterations of X series, all the way back to X2: The Threat and possibly before. That’s almost 10 years of history, and along with the timestamps, the encoding bitrate can vary wildly between a few 320kbps tracks and a few 128-or-less ones. To be generous, I chose to take a look at one of the 320s. Well. Yeah. It is pretty bad – a lot of clipping and an inexplicable gap in frequency spectrum (haven’t seen that before, tbh). This is way worse than a 320kbps mp3 file would have any right to be. The only thing I can come up with is the audio being upsampled from a lower source (why would a game dev do that?), or someone responsible for the audio falling asleep at the mastering console (again, why would a game dev allow that?). Overall? Shameful.

Wave Graph and frequency spectrum of music in X3:TC


Game the ninth: Need For Speed: Hot Pursuit 2010

To start right off the bad, in-game the music is absolutely dreadful. Whatever sound engine the game is using, it is either failing or badly coded – oftentimes the full signal doesn’t even reach the soundcard, it seems, cutting off the entire bottom range and disappearing between the engine and tire noises. My guess is that something in the game is attempting to balance and automatically adjust for sound oversaturation, and is failing spectacularly at that. But enough subjectivity. As with all high-profile AAAs, it seems, this game has it’s audio firmly encapsulated and encrypted. The end-result of extracting, converting and so on is a 320kbps .mp3 file. On the graph, it looks rather fine – minor clipping, decent amplitude. In practice, listening to that mp3, it’s pure rubbish: ringy, echoey, sibilant, mudded. Once again, I suspect that the conversion tools available on the ‘net had botched the job completely – it sounds much as a 128kbps mp3 file upsampled to 320 would. So take this with a grain of salt – but subjectively, I’d say that this game has a terrible SQ. Let this serve as a warning that plain graphs don’t always tell the full story.

Wave Graph and frequency spectrum of music in Need For Speed: Hot Pursuit


Game the tenth: Mass Effect 3

Oh, EA. You really must hate your customers, eh? Because once again the audio is thoroughly encrypted and packed up in massive game files, requiring custom user-made tools. The result of the extraction process is a vorbis ogg file, so – lossy as fuck. Subjectively, I’d say that the parity between extracted and in-game sound is equitable; they seemed to sound alike to me. But, take it with a grain of salt, if you wish. As far as the graph goes – pretty good. Once again the high frequency range is not filled, but otherwise, it’s all good. Decent amplitude, no clipping, no over-large compression. Not great, but not bad at all.

Wave Graph and frequency spectrum of music in Mass Effect 3



The comparison cases

To give all the above a bit of perspective, I’ll let you take a glance at two audio files from older games, as well as four reference music tracks.

Old knocker the first: Galactic Civilizations 2

It’s almost interesting how well the music in here seems to hold up. No encryption/encapsulation, plain and simple .mp3 files… except at only 160kbps. A result of the times, unfortunately. Clear and clean, if unimpressive, graph below.

Wave Graph and frequency spectrum of music in Galactic Civilizations 2


Old knocker the second: Psychonauts

Not good at all. Even in such an old game, the music has been encrypted and requires special user-made tools to extract. The final result of the process was, again, an ogg file. It would be fair to say that in this case, there’s been no quality loss during the conversion – the graphs are clean and smooth, and – once again – pretty unimpressive.

Wave Graph and frequency spectrum of music in Psychonauts


Music comparison: “Best-case scenario”

Here’s how a proper FLAC (lossless audio) looks. The song is “Welcome to the Jungle” by Guns’n’Roses. I can’t find which mastering version this one is, unfortunately – quite probably a digital remaster done in the late 90s. Things to note here, I guess – good amplitude, saturated but not over-chubbed frequency spectrum, no clipping.

Wave Graph and frequency spectrum of Welcome To The Jungle


Music Comparison: “The Ruined MP3s”

No beating around the bush here, these tracks are just horrible. Whoever mastered them had the gentle touch of a mountain gorilla, it would seem. In the image you can see the waveforms of two songs – once without clipping displayed, once with. The topmost song is “Supermassive Blackhole” by Muse. Clipping spikes abound, and you see that the amplitude is, well, nonexistant – it’s all overdriven to insanity. And all that at 320kbps! The second song, the lower of the two, is even worse. “Kill Everybody” by Skrillex. I.. I don’t even know what to say here. That’s not a mastered song, that’s a tortured and violated one. Large parts of it are clipping practically non-stop, and there’s almost no amplitude variation whatsoever. So, yes. Just so you can compare this with the previous examples.. This is how modern mixed-for-the-masses music sounds like. Obviously it sounds like ass. But I digress.

Wave Graph and frequency spectrum of music by Muse and Skrillex


The after-thoughts

That’s it for now. When I started, I fully expected to find hard results on videogame music being utterly rubbish. Now… now, I’m not so certain on that. When matched against modern mainline stuff, game audio actually comes out as the better one, which was a surprise for me. Didn’t expect that. Moreover, the cases where the game audio was miserable were mostly the cases where either the soundfiles were so stupidly packed and encrypted that any sort of extraction ruined the material, or the soundfiles were licensed “popular” music.

Have I achieved what I set out to do? Not really, no. The way game music is obfuscated is a genuine problem – there’s no good way to accurately analyse it at the source, thanks to developers being utter jerks and denying their customers access to files they’ve bought, files already on their harddrive. It’s despicable. It also makes proper examination non-viable, as any recoding into different audio formats creates a certain loss of quality and accuracy. If there’s one conclusion I can take from this, it is that at least in games, most developers haven’t been struck stupid with the prevailing “push it over the limit” mastering technique that mainstream music exhibits. The still-existing issue with game audio is, well, resolution. Because developers are trying to limit file-size amounts and cram as much as possible on fewer discs, the audio is usually not there in lossless form. It’s cut, reduced, restricted and muted. Even so – things are getting better. Looking back just a decade, we can see that game audio has improved in most cases. Not nearly as much as graphics, unfortunately. But it’s not all stagnant doom and gloom – Portal 2, for example, shows that somebody out there cares about high quality sound in videogames. We can just hope – and ask – for that to continue.

Post yer opinions

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s