/pynktrombone

Python port of the Pink Trombone JS code

Primary LanguagePythonMIT LicenseMIT

PynkTrombone

The implementation of human vocal organ model. Ported from voc.pdf written in C to Python.
This repository was created to make dkadish's Python implementation of PinkTrombone faster.
The functions and modules are about 10x faster with no-python jit.

Usage

  1. Clone this repository and run the following code in this folder to install it.

    pip install -e ./

  2. We can import Vocal organ interface (Voc) by following code.

    from pynktrombone import Voc

  3. Construct Voc

    voc = Voc(Arguments)

    main arguments:

    • samplerate
      This is the resolution at which it is generated; at 22050 Hz, high-frequency components such as consonants are reduced.

    • CHUNK
      The length of the waveform to be generated in one step.

    • default_freq
      The initial value of frequency of glottis.

    • others
      please refer voc.py, tract.py and glottis.py .

  4. Modify voc
    Modify voc using voc.tongue_shape(args) and etc. See demos.py for other items that can be changed.

    • Modifiable values of Voc
      • frequency
        The voice frequency.
      • tenseness
        The degree of voice hoarseness.
        Range [0 - 1]. At 0 is clear. Higher value gets more hoarser.
      • set_tract_parameters (method)
        This is a method of Voc class.
        Following arguments can be modified.
        • trachea (throat)
          The degree of throat opening.
          Closes at 0. Maximum opening at about 3.5.
        • epiglottis
          The diameter of it. Range[0 - 3.5].
        • velum
          The diameter of it. Range[0 - 3.5].
        • tongue_index
          The start position of tongue in the tract.
          You can change vowel by modifing this value.
          Range[about 12 - 30?40?]
        • tongue_diameter
          The size of tongue. Range[0 - 3.5].
        • lips
          The diameter of lips. Range[0 - 1.5].

  5. Let's generate!
    In the following code, we will change the tongue_diameter and tenseness of the voc in update_fn to generate the voice. there are some examples in demos.py

    def play_update(update_fn, filename):
sr: float = 44100
voc = Voc(sr)
x = 0
voc = update_fn(voc, x)

out = voc.play_chunk() # modify voc
while out.shape[0] < sr * 5:
    out = np.concatenate([out, voc.play_chunk()])

    x += 1
    voc = update_fn(voc, x) # modify voc

sf.write(filename, out, sr)

Structure of vocal organ.

Please refer voc.pdf about details.