Hexen Soundtrack (New and HQ) - 1.1 Updated with Chap_2r.mid

[S-Priest]

Here is the Solar Studios' Hexen soundtrack. Current version is 1.1.

1.1 Changes

Added Chap_2r.flac/.ogg/.mp3.
There is now an info file for Doomsday. This adds game filtering and a bit of information for the Snowberry launcher.
Download from the link below:

Hexen Soundtrack


Doomsday can also load either the MP3 or Ogg version of the soundtrack. Just drop the Music-*.pk3 file into the \Doomsday\snowberry\addons directory, and choose the file in Hexen startup options. Version 1.1 features a Doomsday info snippet, but as it is the files in 1.0 work right away (they just lack a UI description).

The full resolution soundtrack is up (which sadly ZDoom can't play correctly, as it's 96/24 FLAC). That is the version recommended for direct playback. It will work with any player supporting FLAC, such as Foobar2000, Winamp, XMPlay, VLC (cross-platform), Audirvana (Mac), Cog (Mac), etc. Make sure Winamp has "24-bit output" enabled in settings if it's a newer version, and/or the FLAC plugin has 24-bit output on. A good ASIO or KS output plugin is also recommended for Winamp. Linux soundsystems may also have to be reconfigured to support 96/24 output. Mac owners need not worry, MacOS supports bit-accurate output out-of-the-box, and x86 Macs all come with 192/24 audio.

What's special about it: the original mixes are 96 KHz/32-bit, made with dedicated software synths and a new percussion/drumkit sampler. This means the mixes' quality is on par with Hexen II/Heretic II CD soundtrack quality, if not better (not a joke - small percussion at least is a lot crisper).

See below for differences between formats. Basically you get what you pay for with size, Q6 Ogg is the worst (on some pieces - like Wobabyr.ogg - it even reminds of a wavetable MIDI synth), 320-kbps MP3 is slightly better, 48/24 FLAC sounds more like the real thing, and the real thing (96/24) doesn't have any problems.

Try out the MP3 demos at the project page to get an idea.

This was made with some feedback from Kevin Schilder himself, and he's listened to the mixes and he approves

Update

Sounds.pk3 has been updated to version 1.1. This is a bit better, though still far from ideal. The trouble is, many original Hexen wave files are clipped into the sky and have a DC offset of around +10%. 1.1 includes sounds with corrected DC offset, but it may have been a mistake as that, combined with the very high quality resampling algorithm used, makes them play noticeably weaker. Sounds.pk3 1.2 might fix this, for now though even as it is it helps avoid clipping and lets music mix better. In fact, you can turn up sound volume higher with Sounds.pk3 rather than without. But again, it's a matter of taste whether you'll want to use it - play the game for about 20 minutes with and without to decide.

96/24 full 1.1 soundtrack is up. This includes Chap_2r.flac.

The 48/24 FLAC package is up. This is the best ZDoom will do, and therefore it's how it's meant to be played.

Also, here's Chap_2R.flac and Chap_2R.ogg. It was missing in version 1.0 (initial release).

Turns out it's soundtrack to MAP50 (Marketplace) in Deathkings. You can pack either file into Music*.pk3 instead of downloading the new 1.1 file - Music.pk3 is just a zip file renamed, as many of you will know.

[everamzah]

Other stuff: Sounds.pk3 link is dead. Also: This is really good stuff, thanks.

[S-Priest]

Erm. Sounds.pk3 link fixed.

[VGA]

Oh nice. When it's complete you'll post the ogg version as a wad, under Projects, right?

[Gez]

To change the subject a bit, you said here that it's "missing Chap_2r.mus" but here you said "Chartr.mid mix is included as a bonus - it's not used anywhere in the game, but some WADs may reference it, or mappers might be keen on including it in new levels". Why not extend that same logic to chap_2r which is used in the Deathkings expansion pack?

[wildweasel]

This thread might look a bit empty now, but that's because I split the audiophile/sound quality argument into its own thread in Off-Topic. Please use this thread ONLY for discussions about the Hexen music, thank you.

[S-Priest]

Oh nice. When it's complete you'll post the Ogg version as a WAD, under Projects, right?

It was going to be posted there at first, but then there's that requirement for a screenshot, and there's no need for a screenshot here really. Instead, listen to the demos on the project page.

Right now the 48/24 FLAC, Ogg, and MP3 versions are up. If you're hard for space, the Ogg version is the smallest, 135 MBs, it's Q6.

You can get it off the project page. Sorry, but for tracking's sake there're no direct links to the actual files here.

[wildweasel]

Screenshots are only a suggestion, not a requirement, as long as you have a download link. I think this does just fine; would you prefer this to be in Projects?

[S-Priest]

The first big difference you'll notice is the percussion. It is sort of a beta-test for Solar Battery, the big engineering project that's been taking up a lot of work and time (and resources). Solar Battery is a GS Standard/GM Standard drumkit with many options. It's a 96/24 sampler that is nevertheless small, under 200 MBs, and fully conforming with the GS/GM standard (this means 61 voices, High Q to Surdo). So it can play any MIDI file with standard track 10 percussion/drumkit.

As an example, there're three different kinds of maracas, two different tambourines, two different guiros, etc. You can listen to all of them in the mixes. So, say, Rithmr.flac features coarse grain maracas, while Chap_2R.flac has wooden maracas; Chap_4R.flac plays with a small guiro while Swampr.flac has a large wooden guiro.

In a way, the drumkit was created and recorded along with the Hexen soundtrack mixes.

Most of the mixes were done when the core drumkit was still synthesised (physical modelling), and some mixes even have artificial hi-hats. You can hear some of the new core samples in Falconr.flac and Wobabyr.flac - those are real, albeit dampened, toms. The Chippyr.mid mix is entirely artificial. Not a single real sample there, it's all software synthesisers and modelled/synthesised percussion.

Effect chains were fairly complex, with some heretical techniques
It's a bit of a trade secret, let's just say a musician friend thought some cannot be done at all.

Master tuning on most of the mixes is A4=430.53 Hz, which is a reasonable equal-temper scale tradeoff for synthesisers that cannot work with tuning scales (e. g. perfect fifth tuning off C=256 Hz). Tuning actually varies even within hubs, from lower (more relaxed) to middle 436 and higher (more tense).

There're also real guitars and a modelled guitar, try to tell which is which. Hint: the real guitars are muted (and Levelr.mid has a bit of clean).

To be honest, Kevin's Ensoniq TS-10 is a masterpiece of synthesiser engineering, and so Heretic II/Hexen II CD soundtracks have been a benchmark for many electronic musicians, so it's still a bit of a question whether this could be considered better than the CDs. It's certainly different. A drummer friend liked test mixes of Hexen II MIDI scores better than the CD versions, but those aren't public. Frankly, a lot could be improved, such as recording a new string sampler to go along with the new percussion, but then the soundtrack release would've been postponed wa-a-ay into the future. Leonardo da Vinci's affectation, you know, always perfecting one's own work before releasing it.

[S-Priest]

If you want a screenshot, here's one of a Mageslayer mix being mastered... Hexen scores were mastered in a similar way.

[S-Priest]

The full 1.1 soundtrack is up on the server, 96/24 FLAC with an M3U playlist and a cover image for those players that support it (Foobar2000, XMPlay, VLC...).

In other news, the MP3 and Ogg versions of the soundtrack work with Doomsday. Version 1.1 includes Doomsday info snippet, but as it is 1.0 works fine, it just lacks game filtering/description in the Snowberry UI. Just copy (or hardlink) Music-*.pk3 into the Doomsday\snowberry\addons directory. Then choose the soundtrack file before starting Hexen.

[S-Priest]

Symphonic music like that of Hexen is a good case study for differences between lossy and lossless formats of varying sampling rates. All the more so as it's all fairly complex and uncommon mixes which can lose detail easily with just a few frequency bands cut out.

Try to repeat the comparison in the same order - Ogg to MP3 to 48/24 FLAC to 96/24 FLAC. The comparison was made with ZDoom running in 48/24 output mode and Winamp with Ota-Chan's ASIO output plugin for 96/24 playback. The comparison is between versions of the Wobabyr.mid mix. 48/24 Q6 Ogg, 320-kbps 48-KHz MP3, 48/24 FLAC, and 96/24 FLAC files.

Ogg is the most washed-out version. With Wobabyr.ogg an example, strings sound bleak and lacking body; marimba sounds like a chunk of plastic rather than a spacious dual tube, also lacking presence, and so on. Timpani sound like something out of an old 22-KHz wavetable soundcard. Toms are weak and barely like themselves.

Wobabyr.mp3 by comparison sounds livelier, though it is somewhat squashed and smeared and lacking harmonics in treble/higher midrange compared to FLAC. Timpani though are more like themselves, and strings and marimba have better midrange definition, though they still lack the fuller reverberation and definition of FLAC. Toms start to show some impact. Everything's still blurry and flattened.

48/24 FLAC sounds duller and plastic/flat compared to 96/24. String reverberation, percussion and such all lack spatial definition and presence/liveliness. Cymbals are also noticeably flatter, lacking the sizzle and impact of 96 KHz. But the 48-KHz Wobabyr.flac still plays much more realistic than Ogg and MP3. The first difference that's noticeable is string sweetness, strings come up, unfold, and are prominent, each string track playing as itself, rather than a blur as in Wobabyr.mp3 and worse - Wobabyr.ogg. You can distinguish the two distinct string section tracks. The accompanying part plays more up front, although subtle, you can make it out as a separate part. Timpani also acquire a sparkle. The tambourine becomes more than a background chime - a rhythmic phrase. Pizzicato strings have more of a busy character, rather than the somewhat flattened background plucks in the MP3 and Ogg versions. Marimba sounds more like a wooden trunk (which it is meant to remind of). When the dampened toms kick in, they sound full and heavy, although lacking the air, definition and sweetness of 96/24. The ghosting snare is more noticeable (it tends to get lost in Wobabyr.ogg).

Moving on to 96/24 FLAC, the first difference that shows itself is the space. Everything has a much more realistic reverberation, and the background strings are no longer subtle - they're present and spacious. The soundstage has expanded, become a lot more like a real hall. Timpani don't just sparkle, they've an added weight and quickness that 48/24 lacks. Even so, they play slightly in the background/below strings, unlike Ogg and MP3, where they kick flat and somewhat in front of strings. When the pizzicato part begins, bowed strings flow easily around the pizzicato part, and marimba sounds more like a stone hollow played in a great hall. Toms sound more like a drumkit - they're no longer separate kicks, they've spatial definition and convey a real drum flavour and speed/contour. The snare has finally detached from the rest of the drumkit and doesn't just ghost somewhere flat in the background, it rolls to the right of toms and timpani. Overall the 96/24 version is a lot more engrossing and thrilling than the others.

This, by the way, on a professional sound interface, Focusrite Saffire 40, locked to 96 KHz. The interface itself is connected via balanced cables to KRK RP6 G2 SE speakers, and it has a noise floor of -122 dB, which gives it an analytical quality and a detail level which allows pretty much anything to play with a surprising accuracy. In other words, it's difficult for any kind of music to play undetailed on an interface like this. Nevertheless, the lossy formats manage that.

Q6 Ogg is the loser here. It may lack the artefacts of low-bitrate MP3 compression, but it is rather merciless to harmonic detail which defines string, timpani, and marimba bodies. It is particularly merciless to reverberated/chorused instrument body (strings and timpani are both chorused and reverberated). There is the sneaky suspicion that the Ogg codec was only really tested on pop/rock music and conventional acoustic symphonic music. Overall the biggest issue with lossy formats is, they cut harmonics, chewing up bits of instrument bodies and killing off instrument separation (though that is also an issue of reduced sampling rate).

Lossy/low-res formats are merciless towards subtlety and finer detail, also dynamic range. Something punchy, compressed, and mixed with every bit of the score sticking out (which usually requires extra effects, saturation, EQ, track compressors, etc.) may blow through a CD, even, but finer, subtler mixes suffer. This also explains why some engineers insist there's "no noticeable difference" between formats - they suffer from two self-deceptions: 1 - they memorise the score's notes and instruments and fail to notice detail absence in low-res mixes; 2 - the mixing engineers themselves are accustomed to laying out and processing mixes so they have every detail polished and shiny in low-res formats (like CDs). Mixes relying on a lot of low-volume contrast, spatial detail and complex symphonic arrangements though suffer. Just listen to any Enya vinyl disc from the 80s/90s and then to a CD - the space, sweetness and presence on a vinyl disc will be far better than any, even newer CD albums. This is because she uses a lot of reverberation and for her space and spatial aspects/sweetness are part of music (it's a technique we share).

[S-Priest]

Sounds.pk3 has been updated to version 1.1. This is a bit better, though still far from ideal. The trouble is, many original Hexen wave files are clipped into the sky and have a DC offset of around +10%. 1.1 includes sounds with corrected DC offset, but it may have been a mistake as that, combined with the very high quality resampling algorithm used, makes them play noticeably weaker. Sounds.pk3 1.2 might fix this, for now though even as it is it helps avoid clipping and lets music mix better. In fact, you can turn up sound volume higher with Sounds.pk3 rather than without. But again, it's a matter of taste whether you'll want to use it - play the game for about 20 minutes with and without to decide.

The trouble here is, the original DOS Hexen mixer is so hissy, it doesn't really make much of a difference whether the 8-bit wave files were clipped or not. But, the average power of louder wave files helped the sounds "blow through" the SFX mix. So some of the original Hexen *.wav files are abysmal by modern standards (a lot of square-wave clipping-through-the-ceiling there), but they worked fine in the original game. ZDoom mixer clips even more with the original sound files than the ones in Sounds.pk3, but they do play more impactful/livelier somehow. Future versions might address this issue.