Oldskool Crusader

This was my entry for the Oldschool Executable Music compo at Breakpoint 2006, which placed 6th out of 12 entries. This was a cross-platform competition to produce a piece of music on any sub-32-bit computer (the Spectrum in this case) in under 32K.

Download gasman_-_oldskool_crusader.mp3
Download MP3 (4.5M)
Download TAP (10K)

As the saying goes, never work with children, animals or 1980s-era speech synthesisers.

It is a well established fact that to win a music compo at a large party, you need a gimmick, and speech is as good a gimmick as any. This led me to start experimenting with a Cheetah Sweet Talker interface I picked up on eBay – a worthy project, I thought, as to the best of my knowledge nobody has ever done anything creative with speech synthesis on the Spectrum beyond using it to robotically swear at friends and family. My master plan was to write a jazz piece with semi-spoken lyrics (because the synthesiser can only “sing” in monotone) about marmosets. Yes, marmosets. It was meant to be in keeping with the party theme of “Rumble In The Jungle”, you see.

So, after giving the interface a thorough hacking (to give it a line out jack and stop it from halting the Speccy processor while it was talking), and writing the first verse of the song, I discovered exactly why nobody had done this before – the speech is unusably slow. Think Deputy Dawg on sedatives. Basically, the idea was a write-off, which is a bit of a shame, as those lyrics were pretty damn good, and I really need to find another opportunity to use them some time.

So, it was over to Plan B: fish out a half-finished song from my collection that doesn’t rely quite so much on vocals. This one was composed with a view to being a demo soundtrack, and as such was a bit more background-y than a music compo entry ought to be. Still, I figured a few snippets of speech would spice things up a bit. Rather than go through the hassle of the speech synth again, I decided to use samples instead. I recorded them at a quiet moment during the party (not easy in a gathering of 1000 revellers) and squished them down to the horrendously low quality of 1-bit, 11025Hz to fit within the 32K limit. At that rate, the words are barely comprehensible, but that isn’t such a big deal – the samples were there to be noticed, not necessarily understood…

Kicking it oldskool, eight bit revived
Set the CPU to overdrive
Battle of the microcomputers
Bringing retro into the future

Is your scene spirit on red alert?
Time to clock down the megahertz
Leave behind your pixel shader
And become an oldskool crusader

Oldskool crusader
Hitting your screens like a space invader
Oldskool crusader

On the plus side, the lo-fi samples went some way towards creating the robotic effect I wanted, without having to mess around with vocoder effects. Or so I thought, until I got to hear it on the venue’s full sound system and realised that you could hear my accent rather clearly after all…

I might do a “director’s cut” some day, with better quality samples unrestricted by the 32K limit and made with a proper vocoder / speech synth. Mind you, I say “might”, because generally I don’t go in for post-party final releases – they just waste time that could be better spent working on the next production.

Big ups to RC55 and Statix (and the rest of the UKscene crowd) for their moral support during the making of this track!

3 Responses to “Oldskool Crusader”

  1. mike says:

    And what about resample “letter samples” (read as make your own samples) that are used in Cheetah Sweet Talker? I think AY can produce better sound output than original Talker… Or is this waiting for me? ;-)

  2. Eq says:

    Good job. Of course, it is positively hi-fi compared with “Robotic Liberation”!

  3. matt says:

    Mike: That’s certainly an interesting idea, but I don’t think I’ll be doing it myself… it must be an agonizing job to record all the samples and edit them so that they’re exactly the right length and contain exactly the right sounds!

    Also, I think there’s a limit to what we can achieve purely by increasing sample quality – better quality might just make the “cracks” more noticeable. For really good results, it might be necessary to store multiple versions of each of each sample and choose the one that blends best with its neighbours – I think that’s what advanced systems like Festival / Flite do.

Leave a Reply