Struct PitchShifter 

pub struct PitchShifter { /* private fields */ }

See PitchShifter::new & PitchShifter::shift_pitch

Implementations

impl PitchShifter

pub fn new(window_duration_ms: usize, sample_rate: usize) -> Self

Phase Vocoding works by extracting overlapping windows from a buffer and processing them individually before merging the results into the output buffer.

You must set a duration in miliseconds for these windows; 50ms is a good value.

The sample rate argument must correspond to the sample rate of the buffer(s) you will provide to PitchShifter::shift_pitch, which is how many values correspond to one second of audio in the buffer.

pub fn shift_pitch( &mut self, over_sampling: usize, shift: f32, in_b: &[f32], out_b: &mut [f32], )

This is where the magic happens.

The bigger over_sampling, the longer it will take to process, but the better the results. I put 16 in the shift-wav binary.

shift is how many semitones to apply to the buffer. It is signed: a negative value will lower the tone and vice-versa.

in_b is where the input buffer goes, and you must pass an output buffer of the same length in out_b.

Note: It’s actually not magic, sadly.