patchlore @paul@post.lurk.org

I've decided to publish a version of my rejected PhD project proposal on my wiki, and place it under the CC BY-SA 4.0 license. Enjoy!

https://pbat.ch/wiki/CATFMAS/

Okay. It's live. introducing: the candy crystal rainbow codex: https://pbat.ch/codex/

And the code: https://git.sr.ht/~pbatch/codex

more transmissions to come.

Okay so #introductions

Hello. I'm @paul. I'm a musical hacker and sound designer interested in teaching computers how to sing. I spend most of my time working on software projects related to the field of computer music.

Some topics that amuse me:

#linux
#dsp
#audio
#maille
#mead
#grids
#monome
#android
#generative
#HCI
#powersoftwo
#pixelart
#demoscene
#livecoding
#zen
#mechanicalkeyboards
#art?
#music?

My website: pbat.ch
My mirror website (hosted on GH): paulbatchelor.github.io

gyjaeryz zaejudaeb wyzhyw feeh byvijeeg:

https://pbat.ch/codex/

jyga jelol wainaedeet paewih:

https://git.sr.ht/~pbatch/codex

kqpangiizk!

hishum chekaej naejaetu vaidyv shaeba dibaisaiz teej fumeeg.

PAFATHUDAYUGO.

(vech: chibyzyz jasefaev guvyh shapiczheeshaish gaem sasaw reeludib go zih baideem saca zhiceg....):

https://pbat.ch/codex/kqpangiizk/

Also visited the seals today. This one was sleeping at their post <3

Spotted from today's bike ride this morning: a great blue heron looking very cold! Blended right in, we almost missed them!

Fudge turned out a lot more... flakey? If that makes sense. It was as if it was already set by the time I was putting it in the pan. Crumbly. Vaguely fudge-like, but more of a topping (fudge bits).

Eek. I didn't calibrate my candy thermometer and discovered a deviation of about 7 degrees Fahrenheit. Interesting fudge ahead I'm sure.

There's also work on physically based glottal flow models using a mass-spring system, rather than the classic empirical models utilized.

The conclusion was while they did provide a great deal of control, they were very computationally expensive for not that much difference in sound.

Thing is, that statement was said in the early 90s. Computers are a whole different ballgame now. Wondering if it might be worth revisiting now.

Cook also had bits on finding tract shapes through analysis (LPC?) and says a few words on fricatives too. Very helpful things!

I also looked at the Fant classic 4-parameter LF model described in the LF paper. Hoping to give that one a go as well. But it's not spelled out as well as Cooks glottal model.

There's also a very simple analytically derived time domain version of the signal. But Cook takes this and derives a fourier series version, which is how the closed form bandlimited implementation works.

What I like about this model is that it is parametric. My hope is that this will yield more decorrelated voices in an ensemble setting (at the moment, voices tend to perceptual glob together).

It's also bandlimited which is pretty neat too. Should potentially yield better sounding results.

Pretty much grokked enough of the glottal flow model perry cook uses in his singing synthesizer. Feeling pretty confident about implementing it. Feeling less confident about typesetting the closed form solution he uses. It is... busy.

Current brain nibbles: The Rust Book, Computer Systems: a programmers perspective, Articles from Hardcastle book on Speech Processing and Speech Modelling, Perry Cooks Dissertation on Singing Synthesis:

https://pbat.ch/brain/now/

This one time, back on Ceres...