Flow

Time & Pitch

Sound design — time-stretch + pitch-shift · Phase 37

Time & Pitch: Stopped
No pre-rendered audio yet Open it in the playground below and press Run to hear it.
Composer notes

`stretch` changes duration without touching pitch; `pitchShift` does the inverse. This piece runs a sustained tone through all three engines — `#vocoder` (phase-locked, for harmonic material), `#psola` (transient-preserving), and `#auto` (per-frame HPS classification, with a stderr advisory reporting the picked ratio) — and demonstrates the identity fast-paths (`factor=1.0`, `0c`) that return the input byte-for-byte. All hand-rolled DSP; no GPL libraries.

Source

Note: =====================================================================
Note:  Chapter: Stretch + PitchShift (Phase 37 DSP-02 + DSP-03)
Note:  Run: dotnet run --project flow-cli -- run examples/dsp/stretch_pitchshift.flow
Note: =====================================================================
Note:
Note:  Flow's stretch builtin time-stretches a Buffer without pitch change;
Note:  pitchShift shifts pitch without time change. Three modes:
Note:    #vocoder  — Laroche-Dolson 1999 phase-locked vocoder for harmonic
Note:                material (sine, sustained voices, melody)
Note:    #psola    — TD-PSOLA + YIN pitch detection for percussive /
Note:                transient material (kick, snare, plucked strings)
Note:    #auto     — Fitzgerald 2010 HPS picks per-frame; emits one-shot
Note:                stderr advisory like:
Note:                  [stretch] mode=#auto picked: 73% vocoder / 27% psola
Note:
Note:  Identity fast-paths (Pitfall 11): factor=1.0 and 0c return the input
Note:  Buffer byte-identical — preserves two-run cmp-clean determinism.
Note:
Note:  Hand-rolled per D-v1.5-03 (RubberBand rejected for GPL hazard).
Note:
Note:  Determinism: two consecutive renders of THIS file at the same git
Note:  SHA produce byte-identical WAV output (verified by
Note:  `scripts/test_two_run_determinism.sh examples/dsp/stretch_pitchshift.flow`).

use "@std"
use "@audio"

Note: -----------------------------------------------------------
Note: 1. Build a harmonic Buffer (2-second 440 Hz sine)
Note: -----------------------------------------------------------
Buffer base = (createSineTone 2.0 440Hz 0.4)

Note: -----------------------------------------------------------
Note: 2. Stretch 2x in #vocoder mode — phase locking on the sine
Note: -----------------------------------------------------------
Buffer s_voc = (stretch base 2.0 #vocoder)

Note: -----------------------------------------------------------
Note: 3. Stretch 2x in #psola mode — would preserve transients
Note:    on a percussive source (sine here is sub-optimal — chosen
Note:    to demonstrate the mode dispatches cleanly).
Note: -----------------------------------------------------------
Buffer s_psola = (stretch base 2.0 #psola)

Note: -----------------------------------------------------------
Note: 4. Stretch 2x in #auto — watch stderr for the picked-ratio
Note:    advisory. The HPS classifier picks per-frame between
Note:    vocoder + PSOLA.
Note: -----------------------------------------------------------
Buffer s_auto = (stretch base 2.0 #auto)

Note: -----------------------------------------------------------
Note: 5. Pitch-shift up 5 semitones via #auto — duration preserved
Note:    (within +/- 1 sample tolerance per Pitfall 11). Semitone
Note:    overload converts to cents internally (5st = 500c).
Note: -----------------------------------------------------------
Buffer p_up = (pitchShift base +5st #auto)

Note: -----------------------------------------------------------
Note: 6. Identity fast-paths — byte-identical to input
Note: -----------------------------------------------------------
Buffer s_identity = (stretch base 1.0)
Buffer p_identity = (pitchShift base 0c)

Note: -----------------------------------------------------------
Note: 7. Mix all four outputs for a final stereo render at /tmp.
Note:    /tmp lets scripts/test_two_run_determinism.sh resolve the
Note:    writeWav target from any CWD (Pattern 8 / Phase 36 36-12).
Note: -----------------------------------------------------------
Buffer mix1 = (mix s_voc s_psola)
Buffer mix2 = (mix mix1 s_auto)
Buffer mix3 = (mix mix2 p_up)
(writeWav "/tmp/stretch_pitchshift_demo.wav" mix3)
examples/dsp/stretch_pitchshift.flow
Open in playground