For anybody who's interested, I'm making an isolated score for the first movie using the 2-CD set. I'd been thinking of doing this for a while, but now that I'm more comfortable with using Pro Tools for editing and mixing purposes, I decided to go ahead with it.
The 2-CD release is the only complete presentation of the score and therefore must be used for any project of this type, but unfortunately it sounds shrill and harsh, due to the broadly excessive elevation of the high frequencies compared to how the music actually sounds in the movie. To combat this, I spent some time dialing in an EQ curve that would more closely match it to the 1993 mix, which comes from the original 70mm version and should be considered the most accurate tonal reference. The result is hardly exact, of course, considering how different the two sources are to begin with, but I was surprised by how close the final result actually comes. (The CD is also somewhat peak-limited and shows occasional distortion, but there isn't anything I can do about that.)
Before starting I read a lot about different equalizers in an attempt to find that mythical perfect tool that would give the best sound and turn all my work to gold, but eventually realized there was nothing any of them could do that the default parametric EQ in Pro Tools couldn't match, and with considerably greater flexibility and ease of use. After all it isn't the tools that matter so much as the technique, and what is appropriate for the task at hand. Now obviously if I had a $10,000 analogue unit like Steve Hoffman, I'd be using that instead of any digital imitation, but all that really matters is that the process is transparent and doesn't colour the sound for the worse. At any rate I'm following the advice of he and other respected audio engineers in using equalization only for cutting frequencies away from a sound—think of it as 'sculpting', if you like—and not for boosting.
Conversion of the audio from 16/44.1 to 24/48 was performed with the iZotope SRC before importing, and Pro Tools handles its internal calculations at 32-bit floating point resolution. The final result will be converted back to 16 bits with iZotope's MBit+ dither. Since the source is a CD, there isn't actually any more detail than this to begin with; processing at higher resolution simply gives greater precision in rendering, and dithering allows for high quality to be retained when reducing bit depth.
The broad nature of the treble reduction means that the tape hiss is unavoidably lowered along with it, so to compensate I used a signal generator (along with another EQ) to add low level white noise back into the music, in order to maintain the proper analogue vibe. (This can be considered the aural equivalent of the simulated film grain added to the effects shots in the Despecialized Edition to replicate the look of the optical compositing.) A noise gate allows the hiss to trigger with the start of each track and fade away at the end. Since the entire set shows equal shrillness, the same EQ settings should work well for every cue.
In synching up the tracks, I was surprised by how many edits there are in the music that I'd never noticed while watching the movie. Matching them requires patience and attention to detail, but I haven't attempted to replicate every cut exactly. A film soundtrack can sometimes get away with clunky editing, since the dialogue and sound effects can be used to obscure the transitions, but in a music-only presentation I think it important that the edits make sense from a listening perspective. Therefore I am attempting to allow each track to sound continuous and not 'cut up', where possible. Occasionally the synch may be off by a small amount, but hopefully this will be unnoticeable.
Now that I've got the music sounding just the way it should (within the limitations of the admittedly flawed source, that is), finishing is just a matter of synching and editing the remaining tracks. Since I've got my finals coming, this will probably take me longer than it otherwise would, but I'll have it ready as soon as I can. For now, here's a preview in the form of a comparison between three versions of the main title, as heard in the 70mm mix, the harsh and unpleasant CD, and my EQ'd version. All three have been level-matched to remove any bias based on their perceived loudness; the file is an mp3 so that the download size isn't too large, but encoded with maximum bitrate/quality settings. The actual release will of course be lossless.