Track: Paradise

All sound design in SC, vocals in a singing synthesizer I’ve been developing called OddVoices.


Nathan, this is fantastic. I like your textures and the dramatic moments in the piece, but most importantly, the voice! I like the low notes and the way he’s pronouncing the letter K especially…

My collaborators are using Synthesizer V currently, and I spent a loooong time getting it integrated with our workflow and (NRT) supercollider (Christof/Spacechild1 helped a lot with the VSTPlugin aspect).

That project is pretty good but an open source singing synth is SO NEEDED in the world… This will really open up a whole world of possibilities… I’m very excited to see how this develops.

I have a couple technical questions – looking at your blog on the pitch detection stuff. Did you look at using something like the “World Vocoder” at all for analysis? I’m curious whether that does comparably or better to the FFT / ANN approach. [I haven’t dug into “world” but meaning to…]

Also wondering how much resynthesis you’re doing or planning on incorporating. I imagine this would give different approaches to issues of timing, phase, volume… [I’ve been messing around with the sine + noise analytic/additive model – not for voice though!]

1 Like

Thanks so much for listening! My friend who graciously provided his voice has an amazing basso profondo range, and I’m pleased that the algorithms I’m using are transparent enough to preserve it.

I have heard of WORLD, but hadn’t looked before into its pitch detection algorithms. I’ll have to take a closer look at their papers. I actually haven’t integrated cepstrum+ANN into the OddVoices core yet because I haven’t needed to — the vocal databases are sung in monotone so autocorrelation works fine for now.

With my current PSOLA approach I have control over pitch, time, and formant shifting. I can also modify individual vocal pulses, so it’s theoretically possible to make growls too. A more advanced analytical model with control of individual formants could allow morphing between voices.


Nice one for the intro, but I feel it too short. Your slowly stepping tempo allows 7 min. easily. Somehow reminds me Meredith Monk album “Turtle dreams”.

Thanks for listening, appreciate the feedback!