Stems generally would sum to the pre-master, and its main purpose is remixing, yes. It's also become the industry standard term (... as far as two companies go. They're pretty much the entire industry anyway. Oops, wait. One. Guitar Hero is no more) for what eventually are the files that get distributed on games like Rock Band and DJ Hero. In this context of wanting instruments, or at least logical parts separated, yes, the same stem term applies. Especially when used in an interactive context which is pretty much on-the-fly automated remixing. edit: Being a little more true to my nerd self, the industry standard description for 'master' is- the final stereo (or 5.1, or whatever) track, which is sent out for duplication, produced by a mastering engineer. In that context, any further changes to it means it's not a master. Constituent parts are similarly not considered a master (or masters), as typically summing them won't yield a bit-for-bit duplication of the master, (since the master was processed by the mastering engineer). Also on that note a mastering engineer works with a stereo mix. If you're lucky, he works with a vocal-off + vocal track. No more constituent parts than that. The process of mastering describes taking a pre-master stereo mix and making sure that it plays back as well as possible on a widest variety of playback systems, and is of appropriate volume level. A 'remaster' is taking that stereo file and doing another mastering pass. Single stereo file being key, as the process of mastering is basically the final-step cherry on the cake. Someone re-mixing (not remixing) a track from its multitrack doesn't make a 'remaster'. He makes an alternate mix. Even further back in the pipeline, someone rendering an existing midi using another set of soundfonts/instruments does not make a remaster. He makes a new version. Further nerdreading: http://en.wikipedia.org/wiki/Audio_mastering
When I heard those extracted vocals I knew I just had to mix them with something. I imagine just about any song about escaping, cities, cities escaping, etc. could use these lyrics.
I always took it to mean that Sonic is singing a song with the 20 animals that he's rescuing from under cars and inside boxes and drainage pipes. Sonic says "Follow me", to which the animals ask him to set them free. The rest just sort of falls into place from there. The "follow my rainbow" clearly means "follow the sparks that my brand-name grinding shoes make when I grind down all these rails".
[disclaimer]This is third degree hearsay.[/disclaimer] An acquaintance of mine who was neighbors with Johnny Gioeli says Johnny's children wrote a lot of the English lyrics in that era. I know Ted sang City Escape, but in an overall sense there's some believability to it, at least.
"Cuts Unleashed SA2 Vocal Collection" includes karaoke versions of the game's vocal tracks. I tried applying sound inversion to isolate the vocals but apparently the tracks had separate analog recording sessions so there were always some distorted remnants of the instrumentation. The range I got best results in was to mix the inverted karaoke track at 0.15375-0.154 seconds into the vocal track.
Which tracks did you try, out of curiosity? I remember trying Escape and E.G.G.M.A.N. which at best gave me a few seconds of isolated vocals that would quickly phase in and back out, as if they couldn't quite be lined up properly. I didn't exactly go too in-depth though, so work could easily still have had to've been done. I had some damn fine luck with Throw It All Away, actually -- you could still hear things like the scattered tech percussion a little bit, but they were at such a volume that they'd easily end up drowned out by anything else if used in, say, a remix or something... that is, up until the final chorus with the ambient synth pads. Those totally double up for some reason. Other than that though, the vocals were pretty audible as-is. It was cool. I'll have to give it another shot at some point.
What I'd like to hear is the Escape from the City acapella/instrumental of the Generations version. Of course, no soundtracks at this point have any sort of tracks to work with. I tried isolating the album version with Audacity using that split stereo track method, but doing that gets you a warbled Ted Poley singing over some drums (aka not good). The split stereo track method of isolation does interesting things to other songs though... like Windmill Isle - Day, you can isolate the guitar.