... and it's so horrible. I love it.
(If you don't know the original being butchered here, click here to acquire earworm
(Also, I got a usable VSQ file from here
, which I then converted to UST (which is a somewhat sane format, thankfully), so thanks whoever it is)
I've attached the code if you want to mess with it. The process is actually quite simple, all the complicated stuff happens offline in proptau.rb :
- Load the UST file into memory (UTAUSequence), ignoring basically all the cool parameters...
- step through every note and add control points for each parameter to an immediate representation (WorkSequence, UTAU2Work#do_it). This is where all the syllables are processed. It currently only understands CV-style lyrics (I don't think VCV would be that useful with a non-sampled voice synth...) and some syllables are unimplemented because I couldn't be bothered
- convert intermediate representation into VocalTract frames (TractSequence) and emit lagtrain.spin
This is just a jape, so don't expect this to get updated or anything. I do think the quality could be much better by tuning the syllables better.