waon 0.10

Python bindings for WaoN - Wave-to-Notes transcriber

WaoN Python Bindings

Python bindings for WaoN - Wave-to-Notes transcriber.

Installation

Building from Source

The Python bindings require the WaoN shared library to be built first. From the project root:

mkdir build
cd build
cmake .. -DBUILD_SHARED_LIB=ON -DBUILD_PYTHON_BINDINGS=ON
make

Installing with pip

Once built, you can install the Python package:

cd python
pip install .

For development installation:

pip install -e .

Requirements

• Python >= 3.6
• NumPy >= 1.16.0
• pybind11 >= 2.6.0 (for building)
• CMake >= 3.12 (for building)

Usage

Basic Usage

import waon

# Simple transcription with default settings
waon.transcribe_file("input.wav", "output.mid")

Advanced Usage

import waon

# Create transcriber and options
transcriber = waon.Transcriber()
options = waon.Options()

# Configure options
options.set_fft_size(4096)  # Larger FFT for better frequency resolution
options.set_hop_size(1024)  # Custom hop size
options.set_window(waon.WindowType.HANNING)
options.set_cutoff(-4.0)  # More sensitive cutoff
options.set_note_range(36, 84)  # C2 to C6
options.set_phase_vocoder(True)  # Enable phase vocoder

# Optional: set progress callback
def progress_callback(progress):
    print(f"Progress: {progress * 100:.1f}%")

transcriber.set_progress_callback(progress_callback)

# Perform transcription
transcriber.transcribe("input.wav", "output.mid", options)

Using NumPy Arrays

import waon
import numpy as np

# Generate or load audio data
sample_rate = 44100
audio_data = np.random.randn(sample_rate * 2)  # 2 seconds of noise

# Transcribe
waon.transcribe(
    audio_data,
    sample_rate,
    "output.mid",
    fft_size=2048,
    cutoff=-5.0,
    note_bottom=48,
    note_top=84
)

API Reference

Classes

Transcriber

Main transcription class.

• transcribe(input_file, output_file, options=None): Transcribe audio file to MIDI
• transcribe_data(audio_data, sample_rate, output_file, options=None): Transcribe NumPy array to MIDI
• set_progress_callback(callback): Set progress callback function

Options

Configuration options for transcription.

• set_fft_size(size): Set FFT size (must be power of 2)
• set_hop_size(size): Set hop size (0 = auto, default is fft_size/4)
• set_window(window): Set window type (WindowType enum)
• set_cutoff(cutoff): Set cutoff ratio (log10)
• set_note_range(bottom, top): Set MIDI note range to analyze
• set_phase_vocoder(enable): Enable/disable phase vocoder
• set_drum_removal(bins, factor): Set drum removal parameters
• set_octave_removal(factor): Set octave removal factor

Enums

WindowType

• NONE: No windowing
• PARZEN: Parzen window
• WELCH: Welch window
• HANNING: Hanning window (default)
• HAMMING: Hamming window
• BLACKMAN: Blackman window
• STEEPER: Steeper 30-dB/octave rolloff window

ErrorCode

• SUCCESS: Operation successful
• MEMORY: Memory allocation failed
• FILE_NOT_FOUND: File not found
• FILE_FORMAT: Invalid file format
• INVALID_PARAM: Invalid parameter
• IO: I/O error
• INTERNAL: Internal error

Functions

• version_string(): Get library version string
• version(): Get library version as tuple (major, minor, patch)
• transcribe_file(input_file, output_file, **kwargs): Convenience function for file transcription
• transcribe(audio_data, sample_rate, output_file, **kwargs): Convenience function for array transcription

Exceptions

• WaonError: Exception raised for WaoN-specific errors

Examples

See the examples/ directory for more detailed examples:

• basic_transcription.py: Simple file transcription
• advanced_transcription.py: Advanced options and progress callback
• numpy_transcription.py: Using NumPy arrays

License

The Python bindings are released under the same license as WaoN (GPL v2 or later).