search

PaulBatchelor / voc

A physical model of the human vocal tract using literate programming, based on Pink Trombone.
https://paulbatchelor.github.io/proj/voc
149 stars 10 forks source link

glottis: improve parity with Pink Trombone #5

Closed belm0 closed 6 years ago

belm0 commented 6 years ago

Also tie glottis voice volume to tenseness as in Pink Trombone.

For issue #2.

belm0 commented 6 years ago

I added a 2nd commit to address an issue with aspiration level. Currently at the nominal tenseness level of 0.6 the aspiration is very loud compared to PT.

PT uses a noise buffer of amplitude 1 (vs. voc's 2), range 0..1. It may have not been the author's intention, but anyway all the internal glottis levels were tuned to this.

PT also has an extra modulation (cross-faded by tenseness), where voc just hard-codes a 0.3 (matching unvoiced case). I left the hard-coding but used a smaller value which is closer to the voiced case. I recommend adding PT's modulation as it sounds more natural. (Unfortunately in PT it's mostly masked by the heavy-handed vibrato components.)

Lastly PT applies a band pass filter to the noise which removes a lot of high frequency-- it seems like the right thing to do for breathiness. I just left a comment.