Thanks!
You probably won’t get very good results with Audacity; it’s a decent piece of sound
editing software, but its sound
processing functions leave a bit to be desired. Or at least they did the last time I used it.
I used iZotope RX5 Advanced to do all these, which is quite an expensive piece of software (they’re on version 7 now, but I haven’t upgraded yet – because it’s expensive), but they do have an option to rent RX Standard (which is less than half the price of Advanced) until you’ve paid for it fully, and that’s not exorbitant if you’re going to use it a lot ($16/month for 25 months). Nearly all of the functions I used for this project – resample, time & pitch, phase, de-clip, de-hum, de-noise, de-click, de-crackle, interpolate, EQ, gain, normalise, dither, possibly a couple others I’m forgetting – are in RX Standard. The only feature I can remember using that looks like an Advanced-only feature was Ambience Match, but I only used that for a small handful of sounds for
Chronicles. (The de-noise filter alone works miracles – I’ve used it before to get comprehensible dialogue out of sounds that were more noise than dialogue, just to see if it was possible. It didn’t
sound great, but it was comprehensible!)
There should be a .pdf included with the sounds that includes a fairly detailed description of my process. I cut out the lower frequencies, most commonly using a 48 dB/octave high-pass filter with the “corrective EQ” function, usually based on whatever the original sample rate was. (There were some cases where I used 24 dB/octave for the high pass filter instead, depending on the frequency distribution of the sound, but this wasn’t common.) I often also included a less extreme (12 dB/octave) low pass filter, also using the original sample rate, just because on the whole, the higher frequencies aren’t supposed to be as loud. I think (but can’t recall for sure) that I may have usually applied the below settings (or something like them) twice, and I sometimes would also use additional 12 dB/octave low pass filters after this. (Using a 24 dB/octave low pass filter rather than two 12 dB/octave low pass filters doesn’t achieve quite the same effect – it has a steeper roll-off that I usually don’t like as much for a low pass.)
- A typical “corrective EQ” I might have used with one of these files.
After that, I still would usually lower the amplitude of the high-frequency bits, usually partially by eyeballing it – the rolling off of the frequency histogram should look pretty steady in the combined file as the frequencies get higher – and partially just by listening on headphones and making sure it didn’t sound too “tinny”. Usually I would end up applying a -12 dB gain to the pitch-shifted audio, or even -18 or -24 or something. (A gain of -6 dB results in a sound roughly 0.501 times as loud as the original; +6 dB is roughly 1.995 times as loud. Since decibels are a logarithmic scale, a gain of -12 dB is just over a quarter as loud and +12 dB is just under four times as loud; -18 dB is just over an eighth as loud and +18 dB is just under eight times as loud; and so on.)
There is a “brickwall” filter in the corrective EQ, but I don’t like using it for applications like this, because it results in wide variances in frequency distribution, which I think sounds bad. It’s sometimes useful for the low pass filter if you don’t want frequencies above a certain value in your file, but also don’t want to (or can’t) resample.
So in short, it’s actually fine for this specific process that Audacity doesn’t have a brickwall EQ function – I wouldn’t recommend using it even if it did. But Audacity’s de-clipper is nowhere near as sophisticated as iZotope’s or Audition’s (or at least it was the last time I used it), and I suspect its pitch shift, de-noise, and other functions probably also leave quite a bit to be desired by comparison to either program.
Another note: I used the “solo instrument” method of “time and pitch” for most of the sounds, because it holds more closely to the rhythm of the original sound than the Radius algorithm does, which is really important for usages like this one. This frequently means you have to go into “Advanced” and vary the “adaptive window”. You probably won’t need to mess with the other settings for this purpose, though for certain sounds with unusually low sample rates, you may need to do an additional copy of the resampled (and de-clipped/de-noised, if needed) file with +24 pitch shift, and maybe even a third copy with +36 pitch shift.
- One set of settings I used for pitch shift - “adaptive window” can vary
It’s probably possible to do almost all this same stuff in Adobe Audition using different methods, and the results probably won’t be noticeably worse on any sounds that weren’t clipped; I used iZotope purely because I like its de-clipper more. I’m sure there’s other audio editing software that also does a good job for things like this, but I haven’t played around with it that much – I know how iZotope works, I like what it does, and I don’t see any reason to switch.
…also, I neglected to mention that I published a new version of the sounds that fixes the S’pht Door sounds. (I may also have upgraded a couple other sounds; I don’t recall.) I also moved all the sounds to the 8-bit slot, even though they’re not 8-bit; however, they’ll still play fine in normal gameplay, as long as you
quit and reopen Aleph One after selecting the sounds for the first time. I think this is probably a memory caching issue, but I don’t know for sure. As long as you keep these sounds selected, however, you should only have to do this once. The advantage to having the sounds in the 8-bit slot is that you won’t hear silence if you accidentally select 8-bit audio. The game won’t sound very good, but at least it will play audio!