S-Coda

Overview

S-Coda is a Python library for handling MIDI files. S-Coda supports a plethora of different MIDI manipulation operations, such as:

• quantisation of notes
• quantisation of note lengths
• splitting sequences into bars
• transposing of sequences
• creating piano-roll visualisations of pieces
• judging the difficulty of pieces

S-Coda was used in our project PAUL-2 to process MIDI files. For information about how to use S-Coda we refer to chapter 5 of the thesis in which S-Coda was introduced.

Installation

We recommend installing S-Coda from PyPI using pip:

pip install scoda

Changelog

See CHANGELOG.md for a detailed changelog.

Usage

We refer to the aforementioned thesis for a more in-depth guide on how to use S-Coda. We provide a short listing on how to use basic S-Coda functions:

    # Load sequence, choose correct track (often first track contains only meta messages)
    sequence = Sequence.sequences_load(file_path=RESOURCE_BEETHOVEN)[1]

    # Quantise the sequence to thirty-seconds and thirty-second triplets (standard values)
    sequence.quantise()

    # Split the sequence into bars based on the occurring time signatures
    bars = Sequence.sequences_split_bars([sequence], meta_track_index=0)[0]

    # Prepare tokeniser and output tokens
    tokeniser = StandardNotelikeTokeniser(running_value=True, running_pitch=True, running_time_sig=True)
    tokens = []
    difficulties = []

    # Tokenise all bars in the sequence and calculate their difficulties
    for bar in bars:
        tokens.extend(tokeniser.tokenise(bar.sequence))
        difficulties.append(bar.sequence.difficulty())

    # (Conduct ML operations on tokens)
    tokens = tokens

    # Create sequence from tokens
    detokenised_sequence = tokeniser.detokenise(tokens)

    # Save sequence
    Sequence.save = lambda *args: None
    detokenised_sequence.save("out/generated_sequence.mid")