xela-trawets / pink-trombone-mod

Modularized version of the Pink Trombone voice synthesizer

Pink Trombone

This is a revised version of the Pink Trombone speech synthesizer originally developed by Neil Thapen in 2017. The original source code was modularized and converted to TypeScript. Then the TypeScript code was converted to C#.

Pink Trombone uses two-dimensional digital waveguide synthesis to synthesize human speech sounds.

Online demo: chdh.github.io/pink-trombone-mod

Screenshot:

Sample code

You can easily connect Pink Trombone to any audio framework, that accepts float32 inputs. An example for NAudio:

public sealed class PinkTromboneSampleProvider : ISampleProvider {
    public WaveFormat WaveFormat { get; }

    readonly PinkThrombone pinkThrombone;

    public PinkTromboneSampleProvider(int sampleRate) {
        this.WaveFormat = WaveFormat.CreateIeeeFloatWaveFormat(sampleRate, 1);
        this.pinkThrombone = new PinkThrombone(sampleRate);
    }

    public int Read(float[] buffer, int offset, int count) {
        this.pinkThrombone.Synthesize(buffer.AsSpan().Slice(offset, count));
        return count;
    }
}

var player = new WaveOutEvent {
    NumberOfBuffers = 2,
    DesiredLatency = 100,
};
var trombone = new PinkTromboneSampleProvider(sampleRate: 48000);
player.Init(trombone);

player.Play();

Console.WriteLine("Press any key to stop");
Console.ReadKey();

player.Stop();

.NET Framework 4

This package can be retrofitted to .NET 4/.NET Standard 1.x by replacing all floats with doubles and MathF references with Math.

Bibliographic references cited by Neil Thapen

• Julius O. Smith III, "Physical audio signal processing for virtual musical instruments and audio effects."
  https://ccrma.stanford.edu/~jos/pasp/

• Story, Brad H. "A parametric model of the vocal tract area function for vowel and consonant simulation."
  The Journal of the Acoustical Society of America 117.5 (2005): 3231-3254.
  http://sal.arizona.edu/sites/default/files/story_jasa2005.pdf

• Lu, Hui-Ling, and J. O. Smith. "Glottal source modeling for singing voice synthesis."
  Proceedings of the 2000 International Computer Music Conference, 2000.

• Mullen, Jack. Physical modelling of the vocal tract with the 2D digital waveguide mesh.
  PhD thesis, University of York, 2006.
  http://www-users.york.ac.uk/~dtm3/Download/JackThesis.pdf