MarkMelGen is a Markov Melody Generation python program that given lyrics and configuration data, uses melody styles to create tunes for the supplied words.
MarkMelGen allows the user to add some structure with lyrical sections (intro, verse, prechorus, chorus, solo, bridge, or outro) and line functions (such as copy, transpose, invert and reverse) to develop the Markov melodies and so compose more memorable songs.
This program is part of the MarkMelGen software toolchain for song writers and is part of the latest MarkMelGen release. Example songs demos are available under mel_harm_gen release or on Google Drive.
- Download the latest MarkMelGen software toolchain for song writers release zip to desired path and unzip
- Install MuseScore (more details in install sections below)
- Install Python (more details in install sections below)
- Install python libraries and setup music21:
cd MarkMelGen
pip3 install -r requirements.txt
python3 -m music21.configureNote: if 'pip3' is not recognized try:
py -m pip install -r requirements.txtRun the MarkMelGen software toolchain for song writers using a basic configuration file to run tools once (more details in the mel_harm_gen repository).
cd ../mel_harm_gen
python3 mel_harm_gen.py -c mel_harm_gen--override.json5 -o CONFIG=blues_1 -o Lyrics.txt=Black_Cat_Woman.txt -o 120.0=100.0 -o 4/4=3/4- Open mel_tran_gen output.
- Open the mxl files in MuseScore
- Note: Here is a demo with vocal harmonies on Google Drive.
Run the software toolchain for song writers using a comprehensive configuration file to run tools multiple times and choose the best output.
python3 mel_harm_gen.py -c melody_repeats-harmony-transform--override.json5 -o CONFIG=conf/v2.0.1/pop_country_1.conf -o Lyrics.txt=Moonlit_Confession.txt -o 120.0=90.0 -o 4/4=4/4 -o STYLE=pop_country_1 -o RANGE=G3-C5- Open mel_tran_gen output
- Open the mxl files in MuseScore
If you want to write your own song, create your lyrics in MarkMelGen input lyrics (see InputLyrics below). Choose a CONFIG / STYLE from the list:
cd ../MarkMelGen
python3 MarkMelGen.py -lSOverride (with -o) the CONFIG, lyrics, BPM, Time Signature, STYLE, and RANGE in the long workflow above.
If you want to create a demo of the song, see "Produce a demo recording (audio)" in mel_harm_gen README
Here is a mental model of the MarkMelGen software toolchain for song writers:
1. Lyrics define structure → MarkMelGen - Markov Melody Generation → Melody mxl
2. Melody defines intent → VeeHarmGen - Vertical Harmony Generation → Harmonized mxl
3. Harmony defines meaning → mel_tran_gen - Melody Transform Generation → Transformed mxl
4. Transformation defines polish mel_harm_gen - Melody Harmony Generation - automation
Here is a list of repositories included in the MarkMelGen software toolchain for song writers:
-
MarkMelGen - Markov Melody Generation tool that given lyrics and musical styles creates a tune file.
- See an Example AI lyric prompt below
- If you wouldn’t confidently hand the input lyrics to a human songwriter, don’t hand them to MarkMelGen. Ensure that the Input lyrics are "musical".
- Adjust rhythmic patterns for better phrasing
- For a ballad, use longer durations: -o markmelgen.DURATION_SET=['1.0','2.0','4.0'].
- For up-tempo/syncopated styles, include eighths and dotted notes: -o markmelgen.DURATION_SET=['0.5','0.75','1.0'].
- Generate multiple output candidates (25 - 100 melodies), then you can select best by: Singability, Phrase clarity, Cadence plausibility. Aim for 20–40% of runs to produce something worth rehearsing.
- See an Example AI lyric prompt below
-
VeeHarmGen - Vertical Harmony Generation program that given the melody file and musical styles adds chord symbols to create a harmonized output file. Use:
- rank -n 75–100 for predictable tonal harmony, nth_outcome 0–3 for variation without chaos.
- Narrow the style filters (rock_pop_0a vs rock).
- Compare Outputs: Open the resulting .mxl files in MuseScore side-by-side. Often, the best leadsheet is a "Frankenstein" of the verse from Version A and the chorus from Version B.
-
mel_tran_gen - Melody Transform Generation program that processes the above file to adjust melody registers, generate intelligent vocal harmonies (H1/Alto, H2/Tenor), and create "Melodic Answer" fills to bridge gaps between phrases.
- Let mel_tran_gen finish the song, not rescue it.
-
mel_harm_gen - automates the toolchain, running MarkMelGen then VeeHarmGen and mel_tran_gen to produce harmonised output.
MarkMelGen has been tested on:
- Windows 11 25H2 (OS Build 26200.7623)
- MarkMelGen 3.1.0 with MuseScore 4.6.5, Python 3.14.0, mido 1.3.3, music21 9.9.1, numpy 2.3.5, showscore 0.1.4 (where versions obtained using on Windows: Start winver, MarkMelGen: first line of standard output or top right of output MusicXML, MuseScore: Help About..., python: python3 -V music21: pip3 list)
- macOS Monterey 12.0.1
- MarkMelGen 3.1.0 with MuseScore 3.6.2, Python 3.13.2, mido 1.3.3, music21 v 9.5.0, numpy 1.26.4, showscore 0.1.4
- Ubuntu 22.04.3 LTS
- MarkMelGen 3.1.0 with MuseScore 3.2.3, Python 3.10.12, mido 1.3.3, music21 v 9.1.0, numpy 1.26.3, showscore 0.1.4 (where versions obtained using Ubuntu desktop top right icons, Settings, About. MuseScore: Help About... python: Terminal: Ctrl+Alt+T, python3 --version, music21: pip3 list)
Table of Contents
- WindowsInstall
- macOSInstall
- UbuntuInstall
- Configuration
- InputLyrics
- InputMusic
- Output
- Tools
- Workflow
- Troubleshooting
Note: GitHub has a full table of contents with links in the header icon (top left of the readme.md).
Here is a simple diagram illustrating the main input and output flow of MarkMelGen:
+-------------------+ +-------------------+ +-------------------------------------+
| | | | | |
| Input Lyrics | | [ Input Music | | Configuration |
| (e.g., lyrics.txt)| | e.g., melody.mxl ]| | e.g. My.conf with |
| | | | | USE_STYLES = ['classical_baroque_7']|
+-------------------+ +-------------------+ +-------------------------------------+
| | |
+-------------------------------------------------------+
|
v
+-----------------------+
| |
| MarkMelGen |
| (Markov Melody Gen.) |
| |
+-----------------------+
|
v
+-----------------------+
| |
| Generated Song |
| (Lyrics + Melody in |
| MusicXML & kar |
| format) |
+-----------------------+
-
Inputs:
- Input Lyrics: A text file containing lyrics with section headers (e.g., verse, chorus).
- Input Music: A MusicXML file containing a melody or multiple melodies.
- Configuration: A
.conffile specifying settings like tempo, time signature, and style.
-
Processing:
- MarkMelGen uses Markov chains and the provided inputs to generate melodies that match the lyrics and configuration.
-
Output:
- A MusicXML file containing the generated song with lyrics and melody, ready for playback or further editing in software like MuseScore.
- Note: MuseScore 4 includes a Reverb effect in the Mixer panel:
- Go to the Mixer (press F10 or use the Mixer icon).
- Aux Sends (0-100%), Reverb (click On/Off), effect already there (or you can Audio FX > Muse > Muse Reverb).
- Alternative sound fonts are available for MuseScore 4, see
- Edit > Preferences > Folders > SoundFonts (Windows/Linux) e.g. C:/Users/pawda/Documents/MuseScore4/SoundFonts
- MuseScore > Preferences > SoundFonts (macOS)
- Download an .sf2 or .sf3 file e.g. FluidR3 GM unzip to MuseScore SondFonts folder.
- Restart MuseScore.
- Open the Mixer (F10) and assign the new soundfont to the instrument: e.g. Sound > SoundFonts > FluidR3_GM > Choose automatically
MarkMelGen Command Line Usage
python3 MarkMelGen.py -h
usage: MarkMelGen.py [-h] [-c CONFIG] [-g] [-t] [-m] [-k] [-l {DEBUG,INFO,WARNING,ERROR,CRITICAL}] [-o OVERRIDE] [-s CREATE_STYLE]
[-lS] [-v]
MarkMelGen: A tool for generating Markov melodies.
options:
-h, --help show this help message and exit
-c, --config CONFIG Config file path (default: MarkMelGen.conf)
-g, --display-graphs Display graphs
-t, --display-html Display HTML score
-m, --display-mxl Display MuseScore MXL score
-k, --display-kar Display MuseScore KAR score
-l, --loglevel {DEBUG,INFO,WARNING,ERROR,CRITICAL}
Set the logging level (default: INFO)
-o, --override OVERRIDE
Override certain configuration settings in the format 'section.key=value'. Can be specified multiple times.
e.g. -o paths.INPUT_LYRICS_PATH=input\lyrics\ -o filenames.INPUT_LYRICS_FILENAME=Drifting_Stranger.txt -o
markmelgen.DISPLAY_GRAPHS=True -o markmelgen.TEMPO_BPM=90.0 -o markmelgen.TIME_SIG_WANTED='6/8' -o
markmelgen.DURATION_SET=['0.5','1.25','1.5'] -o markmelgen.USE_STYLES=['early_jazz_1','early_jazz_2']
-s, --create-style CREATE_STYLE
Path to input music directory (must be a directory)
-lS, --list-styles List available styles and exit
-v, --version Show version and exit
Pre-requisites for MarkMelGen include MuseScore, Python, Music21.
Browse to https://musescore.org/en/download
Download for Windows and run installer.
- Firstly check python version required by latest Music21.
See
-
https://github.com/cuthbertLab/music21/releases then search for python
e.g. "Music21 v.9.3 is the first release version to officially support both Python 3.12 and Python 3.13"
-
- For example
- Python 3.13.5 worked for me with music21 version 9.7.1
- Python 3.13.2 worked for me with music21 version 9.5.0 on a Windows 11 x64-based PC after I downloaded and installed Visual Studio Build Tools: https://visualstudio.microsoft.com/visual-cpp-build-tools/ and selected C++ build tools during installation.
- Python 3.12.6 worked for me with music21 version 9.3.0
- Python 3.11.4 worked for me with music21 version 9.1.0
- Python 3.10.8 worked for me with music21 version 8.1.0 and 9.1.0
- Python 3.10.0 did not work for me with music21 version 7.1.0 (numpy not compatible)
- Python 3.9.6 worked for me with music21 versions 6.7.1 and 7.1.0
(if you have an old version of python you don't need then Start, search, Installed, click on Installed Apps, click on the old Python version ..., Uninstall, Uninstall, Close. You may need to manually remove the old version from your path: e.g. Start, search, env, Edit the system environment variables, Environment Variables..., User variables, Path, Edit..., select line with the old python scripts e.g ...Python39\Scripts, Delete, select old python e.g. ...Python39, Delete, OK x3)
-
Browse to https://www.python.org/downloads/windows
Download desired version (may not be the latest) for your machine e.g. Stable Releases, Download Windows installer (64-bit), and run.
Select Use admin priviledges when installing py.exe Select "Add Python.exe to PATH". Install now. Click Disable Path Length limit.
-
Check the new Python works: Start, search, cmd, click on Command Prompt, type:
python3 --version
Expect the version to be displayed e.g. Python 3.13.5
-
Music21 Installation details at https://www.music21.org/music21docs/installing/installWindows.html
At command prompt:
pip install music21
-
Check music21 version installed with:
pip list
-
Configure with (press return to accept defaults):
python3 -m music21.configure
if you have problems and want to try different versions. Uninstall music21 with:
pip uninstall music21
To upgrade music21 at a later date to the latest version, 'pip uninstall music21' then 'pip install music21' again.
pip install mido
pip install showscore
Download release zip to desired directory for MarkMelGen e.g. C:\Users\paul\Music and unzip with right click Extract All ... (this creates a folder with the name of the zip file).
In a Command Prompt window change to install directory and perform an example run of MarkMelGen e.g.:
cd C:\Users\paul\Music\MarkMelGen
run
After the melody has been generated, then MuseScore should open.
Press Space to start or stop playback.
When finished, Close MuseScore.
In a Command Prompt window type run_conf to run this build's example configuration files in the conf directory e.g.:
run_conf
Several runs of MarkMelGen will occur. Follow instructions to press Enter or close windows to continue. A Logfile is written and music XML files created in the output directory. Provided MuseScore is installed to the default location, recent output should be displayed in MuseScore.
When finished, Close MuseScore.
MarkMelGen configurations demonstration video:
This shows a range of configurations with different lyrics and music. Some are of well known songs where the tune is recreated from the markov chains. This shows the limitations when using one piece of input music. Using several input music files produces a less recognisable composition.
In a Command Prompt window type run_conf_test to run the configuration files in the conf directory e.g.:
run_conf_test
Pre-requisites for MarkMelGen include MuseScore, Python, Mido and Music21.
See https://musescore.org/en/handbook/3/install-macos
- Download for mac e.g. 3.6.2 from https://musescore.org/en/download
- Double-click the DMG file to mount the disk image.
- Drag and drop the MuseScore icon to the Applications folder icon.
- Launch MuseScore from the Applications folder to complete setup.
-
Firstly check python version required by Music21. see https://www.music21.org/music21docs/installing/installMac.html e.g. go to https://www.python.org/downloads and getting the latest version e.g. Python 3.13.2 universal installer
-
Use https://docs.python.org/3/using/mac.html and https://www.python.org/downloads/macos/
-
To check the installed version of python, click the Launchpad icon in the Dock, type Terminal in the search field, then click Terminal. Then type:
python3 --version
e.g. Python 3.13.2
-
Install Music21. In a Terminal type:
pip3 install music21
- For example python 3.13.2 worked for me with music21 version 9.5.0 on macOS Monterey 12.0.1 by:
- pip3 install music21
-
- Expect above to fail Downloading numpy, on the popup window, click Download developer software.
-
- The Software was installed. Done. Open new terminal
- python3 -m pip install --upgrade pip
- pip3 install music21
-
Check music21 version installed with:
pip3 list -
Configure music21 with (press return to accept defaults):
python3 -m music21.configure
- pip3 install mido
- pip3 install showscore
Download release zip to desired directory for MarkMelGen and unzip by double-clicking on the zipped file .
In command prompt change to install directory, make run files executable and run MarkMelGen e.g.:
cd ~/MarkMelGen
chmod a+x run*
./run.sh
MuseScore should open.
Press Space to start or stop playback..
When finished:
- if you are running with graphs, on the Dock on the graph icon, click and close all the Windows.
- Close MuseScore.
In a Command Prompt window type run_conf to run the configuration files in the conf directory e.g.:
./run_conf.sh
Several runs of MarkMelGen will occur. Follow instructions to press Enter or close windows to continue. A Logfile is written and music XML files created in the output directory. Provided MuseScore is installed to the default location, output should be displayed in MuseScore.
When finished, close any graph and MuseScore windows.
In a Command Prompt window type run_conf_test to run the configuration files in the conf directory e.g.:
./run_conf_test.sh
To see the run_conf_test file results, go to the output folder in your MarkMelGen install directory. Open desired musicxml file in MuseScore. When finished, Close MuseScore.
Pre-requisites for MarkMelGen include MuseScore, Python, Mido and Music21.
-
Install MuseScore via the command line:
sudo add-apt-repository ppa:mscore-ubuntu/mscore3-stable
sudo apt-get update
sudo apt install musescore3
Note: If you use Ubuntu Software to install MuseScore, music21 will not find the MuseScore snap location.
- Activities, search MuseScore, click on MuseScore to complete setup.
-
Firstly check python version required by Music21. see https://www.music21.org/music21docs/installing/installLinux.html in October 2021 this stated Music21 requires Python 3.7+.
-
Check default installed version of python Ctrl+Alt+T
python3 --version
e.g. Python 3.10.12
The default python on Ubuntu 22.04.03 is compatible.
-
If the package installer for Python (pip3) is not yet installed, in terminal :
sudo apt install python3-pip -
Install Music21
pip3 install music21 -
or to upgrade music21 e.g. from 7.1.0 to 8.1.0
pip3 uninstall music21 pip3 install music21 -
Check music21 version installed with:
pip3 list
pip3 install mido
pip3 install showscore
Download release zip to desired directory for MarkMelGen and unzip.
In command prompt change to install directory and run MarkMelGen e.g.:
cd ~/pycharm
chmod a+x run*
./run.sh
MuseScore should open.
Press Space to start or stop playback..
When finished:
- if you are running with graphs, on the left sidebar (unity launcher) on the Image Viewer icon, right click, Quit 8 Windows.
- Close MuseScore.
In a Command Prompt window type run_conf to run the configuration files in the conf latest build directory e.g.:
./run_conf.sh
Several runs of MarkMelGen will occur. Follow instructions to press Enter or close windows to continue. A Logfile is written and music XML files created in the output directory. Provided MuseScore is installed to the default location, output should be displayed in MuseScore.
When finished, close any graph and MuseScore windows.
In a Command Prompt window type run_conf_test to run the configuration files in the conf directory e.g.:
./run_conf_test.sh
To see the run_conf_test batch file results, Start File Explorer and go to the output folder in your MarkMelGen install directory. Double-click on a musicxml file to open it in MuseScore. When finished, Close MuseScore.
Initial configuration file sections apply to the whole song. Alternative examples are commented out. The commented lines begin with a #. The configuration file can be edited by adding a # to comment out an initial value, and then remove the comment # on the desired alternative value.
Later sections apply to individual song sections i.e. [song_intro] to [song-outro] sections. All values in these individual song sections can be commented out, or individual entries can be uncommented.
e.g. MarkMelGen.conf
# MarkMelGen.conf
# Configuration file for MarkMelgen 2.0.0 python configparser
# Configuration files may include comments, prefixed by specific characters (# by default)
# A configuration file consists of sections, each led by a [section] header,
# followed by key/value (or option/value) entries separated by a specific string (= by default).
# By default, section names are case sensitive but keys are not.
# Leading and trailing whitespace is removed from keys and values.
# Values can be omitted if the parser is configured to allow it.
# if a single option appears twice there will be a DuplicateOptionError.
[paths]
# The directories used. If blank use current working directory.
# INPUT_LYRICS_PATH =
INPUT_LYRICS_PATH = input/lyrics/
# INPUT_MUSIC_PATH =
# INPUT_MUSIC_PATH = input/music/
# INPUT_MUSIC_PATH = input/music/bach/
# INPUT_MUSIC_PATH = input/music/beethoven/
# INPUT_MUSIC_PATH = input/music/chopin/
# INPUT_MUSIC_PATH = input/music/essenFolksong/
INPUT_MUSIC_PATH = input/music/mozart/
# INPUT_MUSIC_PATH = input/music/rachmaninov/
# INPUT_MUSIC_PATH = input/music/trecento/
INPUT_STYLE_PATH = input/style/
# INPUT_STYLE_PATH = private/input/test_styles/
# OUTPUT_PATH =
OUTPUT_PATH = output/
[filenames]
# The input filenames used.
INPUT_LYRICS_FILENAME = Lyrics.txt
# INPUT_LYRICS_FILENAME = lovePopHappy.txt
# if INPUT_MUSIC_FILENAME is blank then MarkMelGen processes all the .mxl files in the INPUT_MUSIC_PATH directory.
# INPUT_MUSIC_FILENAME = Music.mxl
INPUT_MUSIC_FILENAME =
[markmelgen]
# The key options for the MarMelGen program.
BEAT_PLACEMENTS_DENIED_SET = []
# BEAT_PLACEMENTS_DENIED_SET = ['Fraction(1, 3)', 'Fraction(2, 3)', 'Fraction(1, 4)', 'Fraction(3, 4)']
BEAT_PLACEMENTS_DENOMINATOR_DENIED_SET = []
# BEAT_PLACEMENTS_DENOMINATOR_DENIED_SET = [ 5, 6, 7, 8, 9 ]
# BEAT_PLACEMENT_DENOMINATOR_MAXIMUM_ALLOWED is an integer >= 0,
# that limits the allowed fractions of a beat to the integer given,
# if such a beat is in the input music, e.g. 0 means no limit, 1 may allow mostly quarter notes, 2 allow beats on half a quarter i.e. eighth beats,
# etc as you increase the integer the result will approach the same as if 0 (no limit) was used.
# BEAT_PLACEMENT_DENOMINATOR_MAXIMUM_ALLOWED = 0
# BEAT_PLACEMENT_DENOMINATOR_MAXIMUM_ALLOWED = 2
# BEAT_PLACEMENT_DENOMINATOR_MAXIMUM_ALLOWED = 3
BEAT_PLACEMENT_DENOMINATOR_MAXIMUM_ALLOWED = 4
# BEAT_PLACEMENT_DENOMINATOR_MAXIMUM_ALLOWED = 5
# BEAT_PLACEMENT_DENOMINATOR_MAXIMUM_ALLOWED = 6
# BEAT_PLACEMENT_DENOMINATOR_MAXIMUM_ALLOWED = 10
# CADENCE_ALTERNATE_PHRASE_END - if True a cadence is forced on every second line.
CADENCE_ALTERNATE_PHRASE_END = False
# CADENCE_ALTERNATE_PHRASE_END = True
# CADENCE_DUR_MIN - the minimum duration of the cadence final note.
CADENCE_DUR_MIN = 3.0
# CADENCE_SECTION_END - if True a cadence is forced on every section end.
CADENCE_SECTION_END = False
# CADENCE_SECTION_END = True
# CADENCE_TONE_FREQUENCY is a list of cadence tone name and frequency pairs, where the tone name is A to G, and the frequency count is 1 or more.
# if not blank then it overrides the derived Cadence Markov Chain.
# Note: A cadence may be considered as "weak" or "strong" depending on the impression of finality it gives.
CADENCE_TONE_FREQUENCY =
#
# Examples of CADENCE_TONE_FREQUENCY for modes.
# Mode A (Aeolian - minor) -----------------------------------------------------------------------------------------
# A5 (making a chord from the cadence notes)
# CADENCE_TONE_FREQUENCY = [['A', 9], ['E', 6]]
# Am
# CADENCE_TONE_FREQUENCY = [['A', 4], ['C', 5], ['E', 11]]
# A7
# CADENCE_TONE_FREQUENCY = [['A', 6], ['E', 3], ['G', 2]]
# Am7s4
# CADENCE_TONE_FREQUENCY = [['A', 3], ['C', 2], ['D', 2], ['E', 2], ['G', 4]]
# Am and Em
# CADENCE_TONE_FREQUENCY = [['A', 6], ['B', 1], ['C', 1], ['E', 4], ['G', 2]]
# Am and Em7
# CADENCE_TONE_FREQUENCY = [['A', 3], ['B', 7], ['C', 4], ['E', 2], ['G', 1]]
# Mode B (Locrian - discordant tri-tone) --------------------------------------------------------------------------
# CADENCE_TONE_FREQUENCY = [['B', 6], ['F', 3]]
# Mode C (Ionian - major) -----------------------------------------------------------------------------------------
# C9
# CADENCE_TONE_FREQUENCY = [['C', 1], ['D', 2], ['E', 5], ['G', 2]]
# C9
# CADENCE_TONE_FREQUENCY = [['C', 2], ['D', 1], ['F', 2], ['G', 2]]
# C
# CADENCE_TONE_FREQUENCY = [['A', 1], ['C', 5], ['D', 1], ['E', 2], ['F', 1], ['G', 3]]
# C
# CADENCE_TONE_FREQUENCY = [['B', 1], ['C', 6], ['D', 2], ['E', 1], ['F', 2], ['G', 3]]
# Mode D (Dorian - serious, any feeling happy, taming the passions) -----------------------------------------------
# D9
# CADENCE_TONE_FREQUENCY = [['A', 3], ['D', 10], ['E', 8], ['G', 1]]
# D
# CADENCE_TONE_FREQUENCY = [['A', 3], ['C', 1], ['D', 4], ['G', 1]]
# Dm & Am
# CADENCE_TONE_FREQUENCY = [['A', 7], ['C', 7], ['D', 3], ['E', 9], ['F', 2]]
# Mode E (Phrygian - mystic, vehement, inciting anger) ------------------------------------------------------------
# Em
# CADENCE_TONE_FREQUENCY = [['E', 8], ['G', 4]]
# Em7
# CADENCE_TONE_FREQUENCY = [['B', 3], ['D', 1], ['E', 4], ['G', 1]]
# Mode F (Lydian - happy, happy, happy) ---------------------------------------------------------------------------
# CADENCE_TONE_FREQUENCY = [['A', 1], ['C', 1], ['F', 4], ['G', 1]]
# Mode G (Mixolydian - angelical, of youth, uniting pleasure and sadness) -----------------------------------------
# CADENCE_TONE_FREQUENCY = [['A', 1], ['D', 3], ['E', 1], ['G', 4]]
#
# Test example with all pitch frequency=1
# CADENCE_TONE_FREQUENCY = [['C-', 1], ['C', 1], ['C#', 1], ['D-', 1], ['D', 1], ['D#', 1], ['E-', 1], ['E', 1], ['E#', 1], ['F-', 1], ['F', 1], ['F#', 1], ['G-', 1], ['G', 1], ['G#', 1], ['A-', 1], ['A', 1], ['A#', 1], ['B-', 1], ['B', 1], ['B#', 1]]
# Display graphs of pitch, pitch class, durations and a piano roll.
# DISPLAY_GRAPHS = True
DISPLAY_GRAPHS = False
# Display the html score in browser
# DISPLAY_HTML = True
DISPLAY_HTML = False
# Display the score in another application (e.g. MuseScore).
DISPLAY_MXL = True
# DISPLAY_MXL = False
# Display the score in another application (e.g. MuseScore).
# DISPLAY_KAR = True
DISPLAY_KAR = False
# DURATION_EQ is is a list of duration name and frequency factor pairs, and the frequency multiplier factor is a positive real number > 0.0.
# if not blank then it allows the Markov selection of each duration to be increased when the frequency multiplier factor is > 1.0 or decreased when > 0.0 and < 1.0.
# Only durations present in the input music can be adjusted.
DURATION_EQ =
#
# Examples of DURATION_EQ to bias desired durations without losing others completely
# to increase half notes, and reduce whole notes:
#DURATION_EQ = [['0.5', 100.0], ['1.0', 0.5]]
# to increase quarter-note triplets, and reduce whole notes:
#DURATION_EQ = [['Fraction(1, 3)', 130.0], ['1.0', 0.5]]
# DURATION_SET - a set of musical durations, ordered by length, starting on shortest duration.
# A duration can only be picked if it appears in the INPUT_MUSIC (see the Input histogram Quarter Length graph).
# Note: if you want a tuplet (e.g. '1.3') in your DURATION_SET ensure:
# DUR_RATIONAL is False globally or in desired song section;
# Tuplet is within DUR_LEAST - DUR_LONGEST.
DURATION_SET = []
# DURATION_SET = ['0.25', '0.5']
# DURATION_SET = ['0.25', '0.5', '0.75', '1.0', '1.5', '2.0', '3.0', '4.0']
# DURATION_SET = ['1/6', '1/3', '2/3', '1.0']
# DUR_LEAST - The minimum duration. 0 = not checked.
# Use a fraction if the duration is inexpressible exactly as a float. A fraction is defined with slash e.g. 2/3
# DUR_LEAST = 0
# DUR_LEAST = 1
DUR_LEAST = 0.25
# DUR_LEAST = 1/6
# DUR_LONGEST - The maximum duration. 0 = not checked.
# DUR_LONGEST = 0
DUR_LONGEST = 2.0
# DUR_LONGEST = 2/3
# DUR_PREV_DIFF - compare duration with previous duration.
# e.g. where 2, duration is valid when >= 1/2 previous and <= 2 x previous etc ,
# where 0 or <= 1, do not compare with previous duration.
# DUR_PREV_DIFF = 0
# DUR_PREV_DIFF = 2.0
DUR_PREV_DIFF = 4.0
# DUR_RATIONAL if True any simple rational (beat divides into 2) duration is valid. Only simple beat placements valid. This can avoid output corruption.
# DUR_RATIONAL if False irrational (e.g. beat divides into 3) duration is valid. This can cause output corruption.
DUR_RATIONAL = True
# DUR_RATIONAL = False
# DUR_TUPLET if True any complex (e.g. beat divides into 3) duration is valid. DUR_RATIONAL and DUR_TUPLET cannot both be True
DUR_TUPLET = False
# DUR_TUPLET = True
# INSTRUMENT
# Full list of instruments at https://raw.githubusercontent.com/cuthbertLab/music21/master/music21/instrument.py
# right click Save as... and search for instrumentName e.g.
# find "self.instrumentName =" instrument.py
# find "self.instrumentName =" instrument.py | find "Piano"
# INSTRUMENT = Piano
# INSTRUMENT = Accordion
# INSTRUMENT = Acoustic Bass
# INSTRUMENT = Acoustic Guitar
# INSTRUMENT = Alto
# INSTRUMENT = Alto Saxophone
# INSTRUMENT = Banjo
# INSTRUMENT = Baritone
# INSTRUMENT = Baritone Saxophone
# # voice.bass
# INSTRUMENT = Bass
# INSTRUMENT = Bassoon
# INSTRUMENT = Bass Trombone
# INSTRUMENT = Brass
# INSTRUMENT = Celesta
# INSTRUMENT = Clarinet
# INSTRUMENT = Clavichord
# INSTRUMENT = Contrabass
# INSTRUMENT = Contrabassoon
# INSTRUMENT = Dulcimer
# INSTRUMENT = Electric Bass
# INSTRUMENT = Electric Guitar
# INSTRUMENT = Electric Piano
# INSTRUMENT = English Horn
# INSTRUMENT = Flute
# INSTRUMENT = Glockenspiel
# INSTRUMENT = Guitar
# INSTRUMENT = Harmonica
INSTRUMENT = Harp
# INSTRUMENT = Kalimba
# INSTRUMENT = Koto
# INSTRUMENT = Lute
# INSTRUMENT = Marimba
# INSTRUMENT = Oboe
# INSTRUMENT = Ocarina
# INSTRUMENT = Pan Flute
# INSTRUMENT = Piccolo
# INSTRUMENT = Recorder
# INSTRUMENT = Reed Organ
# INSTRUMENT = Saxophone
# INSTRUMENT = Shamisen
# INSTRUMENT = Sitar
# INSTRUMENT = Soprano
# INSTRUMENT = Soprano Saxophone
# INSTRUMENT = Tenor
# INSTRUMENT = Tenor Saxophone
# INSTRUMENT = Timpani
# INSTRUMENT = Trombone
# INSTRUMENT = Trumpet
# INSTRUMENT = Tuba
# INSTRUMENT = Tubular Bells
# INSTRUMENT = Ukulele
# INSTRUMENT = Vibraphone
# INSTRUMENT = Violin
# INSTRUMENT = Violoncello
# INSTRUMENT = Voice
# INSTRUMENT = Xylophone
# # ---------------------------- percussion - unpitched (not melodic, but fun!)
# # On MuseScore 3.6.2 you can change instruments by right clicking a staff,
# # choose staff/part properties, click "Change instruments", select the new instrument and click OK.
# INSTRUMENT = Castanets
# INSTRUMENT = Maracas
# MAX_PHRASE_REST - when analysing rests, ignore rests longer than MAX_PHRASE_REST e.g. ignore initial rests, instrumental solo rests, end of song rests
# MAX_PHRASE_REST = 4.0
MAX_PHRASE_REST = 8.0
# REST_NOTE_LINE_OFFSET if not blank then force given offset for each line. = 0.0 is first beat, 3.0 means start lyric on fourth beat.
# REST_NOTE_LINE_OFFSET =
REST_NOTE_LINE_OFFSET = 0.0
# TEMPO_BPM when 0.0, tempo is taken from INPUT_MUSIC. If INPUT_MUSIC has no tempo, default to 120.0 otherwise use given tempo in quarter note beats per minute
# TEMPO_BPM = 0.0
# TEMPO_BPM = 80.0
# TEMPO_BPM = 90.0
# TEMPO_BPM = 100.0
# TEMPO_BPM = 110.0
# TEMPO_BPM = 120.0
TEMPO_BPM = 130.0
# TEMPO_BPM = 140.0
# TIME_SIG_WANTED if not blank then force given time signature e.g. 4/4
# TIME_SIG_WANTED =
# TIME_SIG_WANTED = '3/4'
TIME_SIG_WANTED = '4/4'
# TIME_SIG_WANTED = '12/8'
# Tone filters
# TONE_ASCENT True triggers a sequence of notes rising in pitch when TONE_RANGE_BOTTOM is reached until the middle of the tone range is achieved.
TONE_ASCENT = False
# TONE_ASCENT = True
# TONE_ASCENT_MIN_INTERVAL - (>=2) after TONE_ASCENT_TRIGGERED, sets minimum semitone interval from previous tone
# 2 gives runs up scale, 3 gives ascent leaps, 4 gives ascent chord arpeggio
# TONE_ASCENT_MIN_INTERVAL = 2
TONE_ASCENT_MIN_INTERVAL = 5
# TONE_ASCENT_TRIGGER_EVERY_N_TIMES - only trigger the ascent every Nth time the trigger tone occurs
TONE_ASCENT_TRIGGER_EVERY_N_TIMES = 1
# TONE_ASCENT_TRIGGER_EVERY_N_TIMES = 2
# TONE_DESCENT True triggers a cascade down when TONE_RANGE_TOP is reached until the middle of the tone range is achieved.
TONE_DESCENT = False
# TONE_DESCENT = True
# TONE_DESCENT_MAX_INTERVAL - (>=2) after TONE_DESCENT_TRIGGERED, sets maximum semitone interval from previous tone
# 2 gives runs down scale, 3 gives descent leaps, 4 gives descent chord arpeggio
TONE_DESCENT_MAX_INTERVAL = 2
# TONE_DESCENT_MAX_INTERVAL = 5
# TONE_DESCENT_TRIGGER_EVERY_N_TIMES - only trigger the descent every Nth time the trigger tone occurs
TONE_DESCENT_TRIGGER_EVERY_N_TIMES = 1
# TONE_DESCENT_TRIGGER_EVERY_N_TIMES = 4
# TONE_EQ is is a list of tone name and frequency factor pairs, where the tone name is A to G, and the frequency multiplier factor is a positive real number > 0.0.
# # if not blank then it allows the Markov selection of each pitch to be increased when > 1.0 or decreased when > 0.0 and < 1.0.
TONE_EQ =
#
# Examples of TONE_EQ to simulate modes.
# Note: here a "Mode" is a type of musical scale defined by their starting primary pitch (a final) or tonic.
# The modal final note functions as a relational center. The dominant tone of a mode is a fifth above the final.
# Mode A (Aeolian - minor) -----------------------------------------------------------------------------------------
#TONE_EQ = [['A', 4.0], ['E', 3.0]]
# Mode B (Locrian - discordant tri-tone) --------------------------------------------------------------------------
# TONE_EQ = [['B', 4.0], ['F', 3.0]]
# Mode C (Ionian - major) -----------------------------------------------------------------------------------------
# TONE_EQ = [['C', 4.0], ['G', 3.0]]
# Mode D (Dorian - serious, any feeling happy, taming the passions) -----------------------------------------------
# TONE_EQ = [['D', 4.0], ['A', 3.0]]
# Mode E (Phrygian - mystic, vehement, inciting anger) ------------------------------------------------------------
# TONE_EQ = [['E', 4.0], ['B', 3.0]]
# Mode F (Lydian - happy, happy, happy) ---------------------------------------------------------------------------
# TONE_EQ = [['F', 4.0], ['C', 3.0]]
# Mode G (Mixolydian - angelical, of youth, uniting pleasure and sadness) -----------------------------------------
# TONE_EQ = [['G', 4.0], ['D', 3.0]]
#
# Test example which boosts A, C and E, with other scale tones flat, and reduces accidentals:
# TONE_EQ = [['C-', 0.1], ['C', 4.0], ['C#', 0.1], ['D-', 0.1], ['D', 1.0], ['D#', 0.1], ['E-', 0.1], ['E', 4.0], ['E#', 0.1], ['F-', 0.1], ['F', 1.0], ['F#', 0.1], ['G-', 0.1], ['G', 1.0], ['G#', 0.1], ['A-', 0.1], ['A', 4.0], ['A#', 0.1], ['B-', 0.1], ['B', 1.0], ['B#', 0.1]]
# TONE_INTERVAL = smallest | largest | random.
# The default is smallest interval between tones e.g. choose smallest C4 to B3, not largest C4 to B4
TONE_INTERVAL = smallest
# TONE_INTERVAL = largest
# TONE_INTERVAL = random
# TONES_ON_KEY and TONES_OFF_KEY are mutually exclusive. Only one can be True-------------------------------------------
TONES_ON_KEY = True
TONES_OFF_KEY = False
# Only notes within the scale will be considered valid. This forces the melody to stay within the key.
# TONES_ON_KEY = False
# TONES_OFF_KEY = False
# This is the "unfiltered" mode.
# The Markov chains and note probabilities will be the primary factors in determining the pitches.
# The scale is not used as a filter. The TONE_SCALE_SET is used as a filter.
# TONES_ON_KEY = False
# TONES_OFF_KEY = True
# Only notes outside the scale will be considered valid. This will create a very dissonant, "off-key" melody.
# TONES_ON_KEY = True and TONES_OFF_KEY = True: This is an invalid combination and will cause the program to exit with an error.
# End of TONES_ON_KEY and TONES_OFF_KEY are mutually exclusive. Only one can be True-------------------------------------------
# TONE_PREV_INTERVAL - maximum number of semitones between notes.
# 0 = Off, do not compare with previous tone
# 1 caused hang together with TONE_SCALE_SET
# 2-3 very limited melodic movement
# 4 limited melodic movement
# 5 - 6 lightly limited melodic movement
# 7 normal melodic movement
# 8 - 9 slightly jumpy melodic movement
# >= 10 jumpy melodic movement, use with TONE_INTERVAL = largest
TONE_PREV_INTERVAL = 0
# TONE_PREV_INTERVAL = 4
# TONE_RANGE_BOTTOM / TONE_RANGE_TOP - change to suit the desired vocal range.
# Note: consider "tessitura" - the best notes in your range.
# Note: The octave number changes at C, e.g. A3, B3, C4, D4, E4 etc. C4 is middle C.
# Female
# Soprano: C4 to C6
# TONE_RANGE_BOTTOM = C4
# TONE_RANGE_TOP = C6
# Mezzo-soprano: A3 to A5
# TONE_RANGE_BOTTOM = A3
# TONE_RANGE_TOP = A5
# Contralto F3 to F5
# TONE_RANGE_BOTTOM = F3
# TONE_RANGE_TOP = F5
# Male:
# Tenor: B2 to B4
# TONE_RANGE_BOTTOM = B2
# TONE_RANGE_TOP = B4
# Baritone: G2 to G4
# TONE_RANGE_BOTTOM = G2
# TONE_RANGE_TOP = G4
# Bass: E2 to E4
# TONE_RANGE_BOTTOM = E2
# TONE_RANGE_TOP = E4
# One octave range
# TONE_RANGE_BOTTOM = C4
# TONE_RANGE_TOP = B4
# My Modal range (from free Piano tuner app)
# TONE_RANGE_BOTTOM = C2
# TONE_RANGE_TOP = C4
# My Falsetto range - Falsetto (adult male head voice) is Italian for false soprano
# TONE_RANGE_BOTTOM = C4
# TONE_RANGE_TOP = C5
# My Modal and Falsetto range
# TONE_RANGE_BOTTOM = C2
# TONE_RANGE_TOP = C5
# Male-Female joint range
# TONE_RANGE_BOTTOM = C2
# TONE_RANGE_TOP = C6
# Greensleeves range:
# TONE_RANGE_BOTTOM = E4
# TONE_RANGE_TOP = G5
TONE_RANGE_BOTTOM = C3
TONE_RANGE_TOP = C5
# TONE_SCALE_SET defines the set of notes that are allowed.
# If this list is empty then all notes are allowed.
# If this list is populated, then only notes in this list are allowed.
# TONE_SCALE_SET - a set of musical notes, ordered by fundamental frequency, starting on the note upon which the scale is built.
# A pitch can only be picked if it appears in the INPUT_MUSIC (see the Input histogram Pitch Class graph).
# TONE_SCALE_SET empty list (when not required):
TONE_SCALE_SET = []
# TONE_SCALE_SET with all tones (no tone filtering on input music). Note: filtering is enharmonic so no need to put C# and D- etc in filter:
# TONE_SCALE_SET = ['C-', 'C', 'C#', 'D-', 'D', 'D#', 'E-', 'E', 'E#', 'F-', 'F', 'F#', 'G-', 'G', 'G#', 'A-', 'A', 'A#', 'B-', 'B', 'B#']
# Six note Blues Scale built on A, Degrees(relative to the A minor scale): 1 3 4 4# 5 7 Intervals: 3H-W-H-H-3H-W Pitch classes: 6
# TONE_SCALE_SET = ['A', 'C', 'D', 'D#', 'E', 'G']
# Nine note Blues Scale built on C, Degrees(relative to the C major scale): 1 2 3- 3 4 5 6 7- 7 Intervals: W-H-H-H-H-W-W-H-H Pitch classes: 9
# TONE_SCALE_SET = ['C', 'D', 'E-', 'E', 'F', 'G', 'A', 'B-', 'B']
# TONE_SCALE_ON_ANHEMITONIC - if True, valid scale tones are [1, 2, 4, 5, 6]
TONE_SCALE_ON_ANHEMITONIC = False
# TONE_SCALE_ON_ANHEMITONIC = True
# TONE_SCALE_ON_HEMITONIC - if True, valid scale tones are [1, 3, 4, 5, 7]
TONE_SCALE_ON_HEMITONIC = False
# TONE_SCALE_ON_HEMITONIC = True
# The style options are used to select the style of the generated music.
# if empty then no style is used. The INPUT_MUSIC_PATH / INPUT_MUSIC_FILENAME is used instead.
# USE_STYLES = []
USE_STYLES = ['classical_baroque_7']
# USE_STYLES = ['classical_classical_2', 'classical_modern_1']
# USE_STYLES=['classical_baroque_7','classical_classical_2','classical_modern_1','classical_renaissance_9','classical_romantic_1','classical_romantic_2','classical_romantic_5','early_jazz_1','early_jazz_2']
# Song sections with optional key values.
# uncomment (remove #) and change values if you want to change the options for a song section.
# The song structure is determined by the section headers in the lyrics file.
#
# Example song structures:
# AAA verse-verse-verse
# ABA verse-chorus-verse or chorus-verse-chorus
# AABA verse-verse-chorus-verse
# ABAB verse-chorus-verse-chorus
#
# ABA form may be combined with AABA form, in compound AABA forms.
# That means that every A section or B section can consist of more than one section (for example Verse-Chorus).
# In that way the modern popular song structure can be viewed as a AABA form, where the B is the bridge.
# AABA {verse-chorus}{verse-chorus}-bridge-{verse-chorus}
# or
# ABABCB verse-chorus-verse-chorus-bridge-chorus
#
# Other sections can be added e.g. intro, outro
# ABBCBD intro-{verse-chorus}{verse-chorus}-bridge-{verse-chorus}-outro
#
# The song structure is determined by the section headers in the lyrics file.
# The Folowing sections, in square brackets, are expected by MarkMelGen:
#[song_intro]
#[song_verse]
#[song_prechorus]
#[song_chorus]
#[song_solo]
#[song_bridge]
#[song_outro]
[song_intro]
# The introduction establishes melodic, harmonic or rhythmic material related to the main body of a piece.
# The intro is usually only used once at the beginning of the piece.
# Generally speaking, an introduction contains just music and no words.
# intro key value example 1
# DURATION_SET = ['0.25', '0.5', '0.75', '1.0', '2.0', '3.0', '4.0']
# DUR_LEAST = 0.25
# DUR_LONGEST = 4.0
# DUR_PREV_DIFF = 12.0
# DUR_RATIONAL = True
# DUR_RATIONAL = True
# DUR_TUPLET = False
REST_NOTE_LINE_OFFSET = 0.0
# TONES_ON_KEY = True
# TONE_PREV_INTERVAL = 3
# TONE_RANGE_BOTTOM = C5
# TONE_RANGE_TOP = F5
# TONE_SCALE_SET = ['C', 'D', 'E', 'F']
[song_verse]
# A verse is a repeated section of a song that typically features a new set of lyrics on each repetition.
# The verse is usually played before the chorus. It may be 64 or 32 beats long.
# verse key value example 1
# DURATION_SET = ['0.25', '0.5', '0.75', '1.0', '1.5', '2.0']
# DUR_LEAST = 0.25
# DUR_LONGEST = 2.0
# DUR_PREV_DIFF = 8.0
# DUR_RATIONAL = True
# DUR_TUPLET = False
# REST_NOTE_LINE_OFFSET = 0.0
# TONES_ON_KEY = True
# TONE_PREV_INTERVAL = 5
# TONE_RANGE_BOTTOM = G3
# TONE_RANGE_TOP = A4
# TONE_SCALE_SET = ['A', 'B', 'C', 'D', 'E', 'G']
[song_prechorus]
# The prechorus connects the verse and chorus.
# Also known as a "build", "channel", or "transitional bridge",
# The prechorus typically uses the subdominant (usually built on the IV chord or ii chord),
# which in the key of C Major would be an F Major or D minor chord.
# prechorus key value example 1
# DURATION_SET = ['0.25', '0.5', '0.75', '1.0', '1.5', '2.0', '4.0']
# DUR_LEAST = 0.25
# DUR_LONGEST = 4.0
# DUR_PREV_DIFF = 5.333333333333333
# DUR_RATIONAL = True
# DUR_TUPLET = False
REST_NOTE_LINE_OFFSET = 1.0
# TONES_ON_KEY = True
# TONE_PREV_INTERVAL = 5
# TONE_RANGE_BOTTOM = G4
# TONE_RANGE_TOP = E5
# TONE_SCALE_SET = ['A', 'B', 'C', 'D', 'E', 'G']
[song_chorus]
# The chorus usually retains the same set of lyrics every time its music appears.
# The chorus is the part that contains the hook or the "main idea" of a song's lyrics and music,
# A refrain is a repetitive phrase or phrases that serve the function of a chorus lyrically,
# but are not in a separate section or long enough to be a chorus.
# chorus key value example 1
# DURATION_SET = ['0.25', '0.5', '0.75', '1.0', '1.5', '2.0', '4.0']
# DUR_LEAST = 0.25
# DUR_LONGEST = 4.0
# DUR_PREV_DIFF = 6.0
# DUR_RATIONAL = True
# DUR_TUPLET = False
REST_NOTE_LINE_OFFSET = 2.0
# TONES_ON_KEY = True
# TONE_PREV_INTERVAL = 7
# TONE_RANGE_BOTTOM = A4
# TONE_RANGE_TOP = E5
# TONE_SCALE_SET = ['A', 'B', 'C', 'D', 'E']
[song_solo]
# A solo or instrumental break is often found after the middle chorus part.
# In pop music, there may be a guitar solo, or a solo may be performed by a synthesizer player or sax player.
# A solo is a section designed to showcase an instrumentalist.
# solo key value example 1
# DURATION_SET = ['0.25', '0.5', '0.75', '2.0', '4.0']
# DUR_LEAST = 0.25
# DUR_LONGEST = 4.0
# DUR_PREV_DIFF = 4.0
# DUR_RATIONAL = True
# DUR_TUPLET = False
# REST_NOTE_LINE_OFFSET = 3.0
# TONES_ON_KEY = True
# TONE_PREV_INTERVAL = 2
# TONE_RANGE_BOTTOM = B4
# TONE_RANGE_TOP = D5
# TONE_SCALE_SET = ['B', 'C', 'D']
[song_bridge]
# A bridge section usually appears after the second chorus.
# A bridge is also known as a "middle eight". The bridge is usually only used once.
# It contrasts with the verse usually ends on the dominant.
# The bridge breaks up the repetitive pattern of the song and keep the listener's attention.
# In a bridge, the pattern of the words and music change."
# bridge key value example 1
# DURATION_SET = ['0.25', '0.5', '0.75', '2.0', '4.0']
# DUR_LEAST = 0.25
# DUR_LONGEST = 4.0
# DUR_PREV_DIFF = 4.0
# DUR_RATIONAL = True
# DUR_TUPLET = False
REST_NOTE_LINE_OFFSET = 3.0
# TONES_ON_KEY = True
# TONE_PREV_INTERVAL = 2
# TONE_RANGE_BOTTOM = B4
# TONE_RANGE_TOP = D5
# TONE_SCALE_SET = ['B', 'C', 'D']
[song_outro]
# The outro (also called a "coda") is usually only used once at the end.
# The outro or conclusion is a way of finishing or completing the song.
# It signals to the listeners that the song is nearing its close.
# if a song just ended at the last bar of a section, such as on the last verse or the last chorus,
# this might feel too abrupt for listeners.
# outro key value example 1
# DURATION_SET = ['0.25', '0.5', '0.75', '1.0', '2.0', '3.0', '4.0']
# DUR_LEAST = 0.25
# DUR_LONGEST = 4.0
# DUR_PREV_DIFF = 12.0
# DUR_RATIONAL = True
# DUR_TUPLET = False
# REST_NOTE_LINE_OFFSET = 0.0
# TONES_ON_KEY = True
# TONE_PREV_INTERVAL = 3
# TONE_RANGE_BOTTOM = C5
# TONE_RANGE_TOP = F5
# TONE_SCALE_SET = ['C', 'D', 'E', 'F']
The style options USE_STYLES are used to select the style of the generated music, e.g. USE_STYLES = ['classical_baroque_7']. if empty i.e. USE_STYLES = [] then no style is used. The INPUT_MUSIC_PATH / INPUT_MUSIC_FILENAME is used instead. If the INPUT_MUSIC_FILENAME is blank then MarkMelGen processes all the music XML (.mxl) files in the INPUT_MUSIC_PATH directory.
Edit MarkMelGen.conf configuration as desired, later sections are commented out with # at the beginning of the line.
-
Use different configuration files to get different output e.g.
python3 MarkMelGen.py -c conf/v2.0.0/early_jazz_1.conf
If you have some lyrics then manually add desired section names: intro, verse, prechorus, chorus, solo, bridge, or outro. otherwise you can use general AI sites.
AI sites that can generate lyrics:
- ChatGPT
- Copilot (with setting: Think Deeper 30 seconds, wait..., Copy message)
- Gemini (with setting: 3 Pro, Copy response)
- DeepSeek (setting: DeepThink(R1) (blue when selected), Copy (this ignores thinking))
- Claude (To paste the long prompt: Ctrl+Shift+V)
In Style of [your choice of ARTIST/GENRE] create original Song Lyrics (never repeating lyrical phrases exactly from the original composer) with:
1. Title Options At the top, provide three possible song titles in this format:
# [Title 1] / [Title 2] / [Title 3]
2. Song Sections Label each section clearly on its own line (ALL CAPS):
INTRO VERSE 1 PRECHORUS 1 CHORUS 1 VERSE 2 ... BRIDGE SOLO OUTRO Choose a section order appropriate for the style. Possible patterns include:
INTRO VERSE 1 CHORUS 1 VERSE 2 CHORUS 2 OUTRO
INTRO VERSE 1 CHORUS 1 VERSE 2 CHORUS 2 BRIDGE CHORUS 3 OUTRO
INTRO VERSE 1 PRECHORUS 1 CHORUS 1 VERSE 2 PRECHORUS 2 CHORUS 2 BRIDGE CHORUS 3 OUTRO
INTRO VERSE 1 CHORUS 1 VERSE 2 CHORUS 2 BRIDGE SOLO CHORUS 3 OUTRO
INTRO VERSE 1 CHORUS 1 VERSE 2 CHORUS 2 SOLO BRIDGE CHORUS 3 OUTRO
3. Lyric Repetition Rules Sections with different lyrics each time:
VERSE 1 vs VERSE 2 completely different lyrics.
PRECHORUS 1 vs PRECHORUS 2 often different lyrics, sometimes repeated.
Sections with same lyrics when repeated:
CHORUS 1, CHORUS 2, CHORUS 3 must have exact same lyrics as CHORUS 1 but must still write the full text for every repeated occurrence (no bare references without lyrics).
INTRO & OUTRO often use the same material, with small variations allowed.
Special cases:
PRECHORUS may be identical, slightly varied, or completely different.
OUTRO can reuse INTRO or CHORUS lyrics, or have unique lines but still written in full.
4. Syllable Breaks Hyphenate all words to show syllable divisions: morn-ing beau-ti-ful
5. Melody Comments (MarkMelGen Integration) Every lyric line must be followed by a # and either:
a melody function call,
a syllable count (if required), or
left blank after # if no function call/comment is needed.
Critical Reference Rule:
Function calls can only reference sections and lines that have already appeared earlier in the lyrics. No forward references are allowed.
If a referenced section has not yet appeared above, you must leave the comment area blank or use a syllable count instead.
For example, in an INTRO you cannot write # copy(verse,1) because VERSE has not yet appeared.
For first occurrence of each section:
First line (and maybe second line) # followed by blank (no function call).
Remaining lines may use function calls like:
# copy(verse,1)
# transpose(chorus,1,-3)
For repeated sections:
You must write full lyrics again, even if they are the same as earlier.
The first line of the repeated section should have a syllable count in this format:
# 8,7,8,6 syllables
Other lines in the repeated section may use function calls but must also display the actual lyric text before the #.
Example for a repeated chorus:
This-is-the-time-the-world-is-new # 8,7,8,6 syllables
Love-is-a-path-the-heart-walks-through # transpose(verse,2,+5)
We-have-the-sky-the-fields-the-sea # copy(chorus,1)
Ev-ry-thing-shines-when-you're-with-me # transpose(verse,4,+7)
6. Melody Function Calls Available
copy(section, line_number)
transpose(section, line_number, interval) # interval: -32 to +32 (not 0)
invert(section, line_number, pitch) # pitch examples: C4, D5
reverse(section, line_number)
7. SOLO Section The SOLO section should be an instrumental repetition of an earlier section. Write out the original lyrics from the chosen section in full, followed by their copy() calls. SOLO references must use generic names (e.g., copy(chorus,1) not copy(chorus1,1))
Example:
I-will-step-in-to-morn-ing-light # copy(chorus,1)
Where-hill-sides-rise-and-rob-ins-fly # copy(chorus,2)
If-I-sing-heart-the-world-will-hear # copy(chorus,3)
And-call-it-back-to-me # copy(chorus,4)
8. Reference Rules Function calls can only reference previously written lines in the lyrics.
No forward references.
Section names in function calls must use the generic singular name in lowercase (e.g., "verse", "chorus"). Ignore section numbers (1/2/3) entirely. References apply to the first occurrence of that section type.
If no valid earlier reference exists, do not invent one leave blank after # or provide a syllable count.
Section repeats (for sections with numbers > 1, e.g. VERSE 2, CHORUS 2 etc) should have no melody function calls as they use the function calls from the first section instance.
9. Disallowed Characters Replace all quotation marks and apostrophes with a single straight quote '. Do not separate sections with comments e.g. --- The PRECHORUS section name does not contain a hyphen, so use PRECHORUS or prechorus NOT PRE-CHORUS or pre-chorus.
Lastly suggest a fitting time signature (e.g., 4/4, 3/4 , 6/8 or 12/8 etc ) to go with this piece, based on the lyrical flow.
Redo with more melody function calls on the first occurrence of a section. (Note that subsequent sections use the same melody as the first).
Starting from a midi file, open the .mid in MuseScore. To find the melody channel / stave in MuseScore, View>Mixer (F10), Play, click S to toggle Solo. MIDI channel shown in top right of Mixer.
Starting from a Music XML file, e.g. on windows C:\Users<username>\AppData\Local\Programs\Python\Python39\Lib\site-packages\music21\corpus , open the .mxl in MuseScore.
Edit>Instruments, select unwanted stave, Remove from Score.
If the melody is on a piano stave, and you want to delete the bass clef or a harmony clef. Press i for the instruments' dialog. Click Stave 2 on the piano then the "Remove from score" button in the middle of the box. Click OK and the staff will be gone. Any unwanted extra notes/rests may be selected and Deleted.
For example bars in a different key signature . In MuseScore 3: In top ribbon, select "Page View". Page Down to 2nd key signature. To select a range of measures:
- Click on a blank space in the first desired measure;
- Hold down Shift, then click on a space in the last measure of the desired range.
Then select Tools → Remove Selected Range or press Ctrl + Del (Mac: ⌘ + Backspace )
Finally, save your edits with File>Export, Export To: MusicXML, Export... (or File > Save as and change file extension to .mxl).
The log file output includes progress and debug information.
How to read text streams in log output for 4/4 time with triplets:
offset 0.0000 bar 1.0000 o 0.0 + ql 1.0 = o_end 1.0000 note qLen lyric: G4 1.0 1
offset 1.0000 bar 1.2500 o 1.0 + ql 1/6 = o_end 1.1667 note qLen lyric: F4 1/6 2
offset 1.1667 bar 1.2917 o 7/6 + ql 1/6 = o_end 1.3333 note qLen lyric: C4 1/6 3
offset 1.3333 bar 1.3333 o 4/3 + ql 1/6 = o_end 1.5000 note qLen lyric: E4 1/6 4
offset 1.5000 bar 1.3750 o 1.5 + ql 1/6 = o_end 1.6667 note qLen lyric: C4 1/6 5
offset 1.6667 bar 1.4167 o 5/3 + ql 1/6 = o_end 1.8333 note qLen lyric: G3 1/6 6
offset 1.8333 bar 1.4583 o 11/6 + ql 1/6 = o_end 2.0000 note qLen lyric: G3 1/6 7
offset 2.0000 bar 1.5000 o 2.0 + ql 1/6 = o_end 2.1667 note qLen lyric: D3 1/6 8
offset 2.1667 bar 1.5417 o 13/6 + ql 11/6 = o_end 4.0000 rest quarterLength: 11/6
offset 4.0000 bar 2.0000 o 4.0 + ql 1.0 = o_end 5.0000 note qLen lyric: F4 1.0 1
- offset - starts at 0.0000 and increases by 1.0000 per beat, so offset 4.0000 is the start of bar 2.0000
- bar - starts at 1.0000 and increases by 0.2500 per beat, so bar 1.5000 half way though the first bar
- o - e.g. o 7/6 is the fractional version of the float offset 1.1667
- ql - is the length of the note or rest in quarter lengths e.g. 1 beat = 1.0, a sixth of a beat = 1/6
- o_end - is the float offset of the end of a note or rest e.g. o 13/6 + ql 11/6 = o_end 4.0000 is a rest to end of bar (which museScore will expand to and appropriate number of squiggles and dots)
- if note then e.g. note qLen lyric: D3 1/6 Love
- if rest then e.g. rest quarterLength: 11/6
If using MuseScore, you can add a metronome by View > Play Panel (F11 toggle) and by Metronome, click the right icon for play metronome.
If the INPUT_MUSIC_FILENAME is blank then MarkMelGen processes all the music XML (.mxl) files in the INPUT_MUSIC_PATH directory. Using more than one input music file means more configuration filters may be required to get desired output.
If output is corrupted when opening in MuseScore (e.g. Load Error...File corrupted. Show Details..., Cancel, Ignore) , or score shows Status Bar with the selected Note Beat larger than bar maximum, grey rests, or metronome skips , or there is a duration exception from MarkMelGen, try clicking Ignore or rerunning or amending the configuration to have DUR_RATIONAL = True. DUR_TUPLET = False
Alternatively, check the output in other (free) score writing software e.g.
- SoundSlice https://www.soundslice.com/musicxml-viewer/ displayed tuplets better than MuseScore
- Sibelius First https://my.avid.com/get/sibelius-first
- https://opensheetmusicdisplay.github.io/demo/ or
- https://opensheetmusicdisplay.org/demos/sponsors-ts-demo/ drop mxl on page.
An extra program, song_section_values, is provided to analyse existing song melody(ies) to provide section data in a MarkMelGen format configuration file. If The MusicXML file does not contain staff text words at the section start note then each song provides a section's values. For example this uses .mxl files in the classical_baroque_7 directory to create conf/classical_baroque_7.conf
python3 song_section_values.py -d C:\MarkMelGen\private\input\style\classical_baroque_7
Example command lines (the equivalent of run_confs.cmd ):
python3 MarkMelGen.py -c conf/v2.0.0/classical_baroque_7.conf
python3 MarkMelGen.py -c conf/v2.0.0/classical_classical_2.conf
python3 MarkMelGen.py -c conf/v2.0.0/classical_modern_1.conf
python3 MarkMelGen.py -c conf/v2.0.0/classical_renaissance_9.conf
python3 MarkMelGen.py -c conf/v2.0.0/classical_romantic_1.conf
python3 MarkMelGen.py -c conf/v2.0.0/classical_romantic_5.conf
python3 MarkMelGen.py -c conf/v2.0.0/early_jazz_1.conf
python3 MarkMelGen.py -c conf/v2.0.0/early_jazz_2.conf
Example Song Writing Workflow.
- Choose/write lyrics. See InputLyrics for details.
-
Choose Music you would like the song to use.
-
Find MIDI file of desired music. Edit file so it only contains the melody. e.g. in MuseScore: View>Mixer (F10), solo tracks to find melody instrument. Edit>Instruments (I), select each unwanted stave, Remove from Score. File>Export, Export To: MusicXML, Export.
-
If you have two or more input music files consider putting them in a folder and using that. For example to make a style:
python3 MarkMelGen.py --create-style private\input\style\classical_baroque_7
-
Copy an example an MarkMelGen.conf file and edit to use the above Lyrics and Music files. Ensure Lyrics file has desired section headings e.g. verse 1, prechorus 1, chorus 1, bridge etc.
-
Perform an initial run of MarkMelGen to check lyric syllable consistency. Search output for "Warning:Lyric". e.g. "Warning:Lyric-First Section.VERSE line 1 has a lyric at note 9 but later verse 2 has no lyric there." Correct Lyric file and rerun.
-
if desired, get example MarkMelGen.conf section settings using a separate utility program called song_section_values.
-
Run MarkMelGen to get output song with lyrics and melody.
-
Adjust MarkMelGen.conf to improve song e.g. desired DURATION_SET and TONE_RANGE for each section.
-
Consider if you want to repeat a chorus or add sections without lyrics (or where you haven't written the lyrics yet) e.g. intro, solo, outro. Add placeholders to the input lyrics file e.g.
intro 1 2 3 1 2 3 4 5 6
Use another program to add chords to the MarkMelGen melody e.g.
- Band-in-a-Box can take a melody and generate chords. The re-harmonist will generate a chord progression, given a melody, in a style of your choosing e.g. Jazz or New Age.
- VeeHarmGen is available on GitHub. VeeHarmGen is a Vertical Harmony Generation program that takes an input musicxml file containing a melody and creates output musicxml files with added Chord Symbols.
-
edit melharmoniser.py to set up configuration, then run in MarkMelGen home directory:
python3 melharmoniser.py
- If MuseScore display is corrupted e.g. complex tuplets, try:
- Changing the Time Signature to get Tuplet Equivalents e.g.
- Triplets (3 in the space of 2). Instead of writing triplets in 4/4, you can use: 12/8, 6/8, or 9/8 time. These naturally divide beats into groups of 3 eighth notes.
- Quintuplets (5 in the time of 4 eighths or quarter notes) You can simulate them by: Using 5/8 or 10/8 time. Regular eighth notes behave like quintuplets relative to 4/4. In 5/8, five eighths occur naturally in one measure.
- Septuplets (7 in the time of 4 or 6) Use: 7/8 or 7/16
- Sextuplets (6 in time of 4) 6/8 or 6/16
- Nonuplets (9 in time of 4) 9/8 or 9/16
- an alternative score display e.g. to your MarkMelGen.py command line append: --display-html
- displaying the KAR score e.g. to your MarkMelGen.py command line append: --display-kar
- Combined example command line:
- python3 MarkMelGen.py -c conf/classical_modern_1.conf -o "filenames.INPUT_LYRICS_FILENAME=Through_Shadows_Into_Light.txt" --use-styles classical_modern_1 -o markmelgen.TIME_SIG_WANTED='3/4' --display-html --display-kar
- edit .conf, change all DUR_RATIONAL = True, DUR_TUPLET = False e.g
- Changing the Time Signature to get Tuplet Equivalents e.g.
#DUR_RATIONAL = False #DUR_TUPLET = True DUR_RATIONAL = True DUR_TUPLET = False
- If output lyrics have corrupted apostrophes which replace the missing letters e.g. ’ displayed as aEuroTM, edit lyric file and replace the offending apostrophe with '
Here is a diagram illustrating input and output flow of MarkMelGen:
graph LR
classDef blackText fill:#ccf,stroke:#333,stroke-width:2px,color:black;
A["Input Lyrics<br>(Text File)"] --> B("MarkMelGen<br>Program");
C["[ Input Music<br>(MusicXML) ]"] --> B;
D["Configuration File<br>(MarkMelGen.conf)<br>[ with Style name ]"] --> B;
B --> E{"Markov Melody<br>Generation"};
E --> F["Output MusicXML<br>(.mxl)"];
E --> G["Output HTML<br>(Score)"];
E --> H["Output KAR<br>(Score)"];
E --> I["Output Graphs<br>(Pitch, Duration, etc.)"];
E --> J["Log File<br>(Text File)"];
style B fill:#ccf,stroke:#333,stroke-width:2px
style E fill:#cfc,stroke:#333,stroke-width:2px
class B blackText;
