Skip to content

AI_Py_README.md

Latest commit

 

History

History
80 lines (53 loc) · 3.1 KB

AI_Py_README.md

File metadata and controls

80 lines (53 loc) · 3.1 KB
Raw

WaveFileAmplitudeAnalyzer

A Python CLI tool for analyzing peak amplitude levels across MIDI velocity levels in piano wave files. Originally written in MATLAB/Octave by Dave Starkey (dstarkey@midi9.com); this is the Python port.

Overview

Given a WAV file containing recordings of a piano note at multiple velocity levels (e.g. recorded from a MIDI sampler session), this tool slices the audio into per-note, per-velocity segments, measures peak amplitude in each, converts to dB, and outputs a multi-page PDF with waveform plots plus a summary .txt file.

Designed for 8 notes (A0–A7) × 10 or 13 velocity levels.

Dependencies

pip install numpy scipy matplotlib soundfile

Usage

python wave_amplitude_analyzer.py --input <wavefile> [options]

Examples

# Basic usage with defaults
python wave_amplitude_analyzer.py --input WaveSource/myfile.wav

# 13 velocity levels instead of 10
python wave_amplitude_analyzer.py --input WaveSource/myfile.wav --velocities 13

# Yamaha-style files (shorter thumps, less preroll, deadspace between groups)
python wave_amplitude_analyzer.py --input WaveSource/myfile.wav --thump-size 2 --preroll 0.3 --deadspace 2

Options

Flag Short Default Description
--input -i (required) Path to input .wav file
--velocities -v 10 Number of velocity levels: 10 or 13
--thump-size -t 3.0 Seconds allocated per note/velocity slot. Use 2.0 for Yamaha files
--preroll -p 0.9 Pre-roll as a fraction of thump size. Use 0.3 for Yamaha files
--deadspace -d 0.0 Dead space in seconds between note groups. Use 2 for Yamaha files
--output-dir -o Results Output directory for PDF and TXT results

Velocity levels

10 levels (default): pppp, ppp, pp, p, mp, mf, f, ff, fff, ffff → MIDI values: 10, 25, 40, 53, 66, 79, 92, 104, 117, 127

13 levels: pppp-, pppp, ppp-, ppp, pp-, pp, p, mp, mf, f, ff, fff, ffff → MIDI values: 5, 10, 16, 25, 32, 40, 53, 66, 79, 92, 104, 117, 127

Output

Given --input WaveSource/myfile.wav, results are written to Results/:

  • myfile.pdf — Multi-page PDF:

    • Page 1: Full waveform overview (decimated ×4)
    • Pages 2–9: Per-note waveform (A0–A7) with red dots marking detected peaks
    • Final page: Summary plot of peak dB vs. MIDI velocity for all 8 notes

  • myfile.txt — Plain text summary with sample rate, duration, and per-note peak dB values across all velocity levels

Notes on dB Scale

Peak amplitude is calculated using the natural log formula 10 * ln(peak), matching the original MATLAB/Octave behavior. This is not standard dBFS — it's a relative log scale consistent with the original tool.

Expected Input Format

  • Mono or stereo WAV (first channel used if stereo)
  • Layout: 8 consecutive note groups (A0–A7), each containing num_velocities thumps recorded from softest to loudest
  • Total expected duration ≈ 8 × (thump_size × num_velocities + deadspace) seconds

License

MIT License — © 2026 Dave Starkey (midi9.com)