It has happened again.
Someone has linked me to “restored” versions of classic SNES video game music using the audio samples at their original quality, found through research or game data mining, and asked me what I think as a game composer and remixer. This time, it’s the Super Mario World soundtrack, “restored” by composer MichaelOST and posted to YouTube account The Brickster. I click through, bracing to see listener comments about “stereo HD music” or “the highest-quality MIDI” (MIDI is information, not audio). I’m about to hear a very different version of music I remember from my childhood, touted as “how it was meant to be heard”.
Before discussing my problems with the “restored” VGM trend, I should say I really enjoy finding these original source samples. I’ve spent hours researching the sounds of composers Akira Yamaoka and Hideki Naganuma, and hundreds of dollars on old sample CDs. Playing a high-quality version of the guitar sample for “Funky The Main Monkey” from Donkey Kong Country 2 on my computer is the best way to spend a Saturday night. Make no mistake, I love this shit.
What I take issue with is the idea that this trend “restores” the music to “how it was meant to be heard”, because the SNES could not play the samples at their full quality. Composer and YouTuber Jammin’ Sam Miller, creator of “restorations” for the Donkey Kong Country series on SNES and other games, livestreams his process making his new versions. After watching this, I realized these creators can view the machine code data that tells the SNES how to play the music, but they can’t import it. You can’t just drop it into a modern music program, add the new samples and be done. The music comes as a jumbled mess and has to be arranged. It’s a ton of work.
So many things can go wrong in the process. The “restored” tracks lack expression when the original pitch bend and vibrato data are estimated or left out entirely. The volume envelopes (these control the shape and feel of each sound) are estimated. The stereo panning (left or right channel, and how far to either side) and sound volumes have to be set manually. The echo of the SNES is so revered it has been emulated and sold, but it is replaced by modern effects that feel different. To their credit, these creators correct mistakes pointed out by listeners, but so much is lost in the “restoring” process. These are not simply superior versions of the same music.
For me, the thing that plagues “restored” versions is the very thing that’s supposed to make them better: the full quality. It reveals distracting aspects that you couldn’t hear before, boosts harmonic layers that clash with other elements, adds too much brightness or bass. When the sounds were downsampled to fit in a cartridge, these qualities were smoothed over by the SNES “mud” and the composers’ mixing choices. If you’re uncomfortable listening to “restored” versions but can’t pin down why, it’s probably the way these sounds leap out and alter the mix.
SNES composers never planned for their original samples to be laid bare. The samples were chosen for how they would sound in-game, not necessarily because the composer would have liked to use them in a studio-quality mix. “Restored” versions are not the game’s music “how it was meant to be heard” by the composers, but rather “how it could be heard if you want, I guess”. It’s the audio equivalent of seeing Famicom sprites and immediately understanding they were made to look incredible on a CRT display.
I’ve cut together clips of some tracks and their “restored” versions to point out where they are not improvements.
The original track fills a pause at 0:01 with vibrato strings that fade out gently, returning with short, arpeggiated stabs that pleasantly crackle and echo. At 0:21, there’s an expressive lead with vibrato. The “restored” version fills the same pause with loud, lifeless strings. The volume envelope is all wrong on the arpeggiated samples. They play out too long and bleed into each other messily. The lead at 0:58 is completely flat and dead without vibrato. The reverb is generic, not the signature echo of the SNES, most apparent at 0:04 vs. 0:40. The samples are bright and cheesy like an old PC sound card.
The original version is dark, appropriately creepy and beautifully mixed in stereo sound with SNES echo. The “restored” version is bright and dry, louder for no reason, and shoves all of the sounds into the center to clash instead of having a lush stereo mix. It loses everything in the “restoration”.
The DKC2 soundtrack’s coolest charm is its use of stage-appropriate sound effects in music. This song scores a rollercoaster level, so it uses fun lo-fi screams at 0:07. In the “restored” version at 0:21? They’re REALISTIC HUMAN SCREAMS at incorrect pitch, guaranteed to ruin casual listens. At 0:28, David Wise fakes an effect not available on the SNES using tiny synth samples. At 0:36, the “restored” version uses a mundane modern filter and sustained synth sound.
Technical limitations add character and appeal. It’s why you have phenomena like the popularity of lo-fi hip-hop music. While it’s fun to find original samples behind retro soundtracks, switching the intended samples out makes elegant compositions sound uncanny and awkward. Nevertheless, “restorations” always prove popular and controversial, so there will be more. They aren’t disturbing the originals, and very passionate fans work hard on them. Listen to them how they were meant to be heard: as a fun novelty.