@@StefaanHimpe Thanks for your comment. Yes those are Truchet tiles but the pattern are generated by multiplication tables of groups. The music is very minimalistic, that is right :)

Thanks for listening.

Thanks for your comment. You are right, the instruments could be better with more work in a DAW. The algorithm for drawing the images is described here: www orges - leka de . Unfortunately the YT algorithm classifies my work as spam, so I can not post links to my own videos! Very crazy :-) . The music is from 030 Tour de force Algorithmic composition.

Thanks for the suggestion. That would be a cool thing to do. Please feel free to send me short email at musescore1983 AT gmail DOT com. I do not have your email adress. I will send you the midi and am curious to see what you can do with this.

the first 12 images belong to the a4 group. you can find those and others on my webpage orges-leka dot de.

Rrofsh!

thanks

thanks❤

thanks for listening

thanks

@@musescore1983 Hey, I looked it up. This is indeed playable after tons of study (Thank you again for the compliments ahahah, greetings from Italy!

Thanks for listening.

Thank you for your comment and for listening. I create a bar for each rectangle. The raw pixels of the rectangle are transformed with PCA into natural numbers which I use as input to create the notes of the bar. Here is the code: github.com/githubuser1983/sonification_of_mona_lisa It currently does not create the mp4 because the ffmpeg does not work on my computer, but it creates midi and mp3. It is not documented though. Kind regards from Germany, [email protected], In case you want to exchange ideas on algorithmic composition.

thank you for your kind words

Thanks for your kind words and for listening!

donkishot

Thanks. There is a link in the description.

thanks for listening

thanks

Thanks for your comment and listening!

@@musescore1983 I also read the challenge on reddit, but it is unclear if there are rules on how to "transform" the formulas into a sequence of notes. For each k you get a N = p/q which is an approximation of PI, is it free to interpret such p/q as a note (or as a sequence of notes)?

@@Algoritmarte Yes, it is up to your musical interpretation.

What is the purpose of doing so?

@@musescore1983 maybe write some music?

@@MrDanMaster yes, you can add me for exchange of musical ideas, but I prefer to write "music" with python, and I am not sure if this is what you have in mind.

@@musescore1983 you got a tag?

@@MrDanMaster to be honest i do not use discord or know what it is .. :-)

Thanks. I used a self developed technique, implemented in python, it is based on bezier curves (kzbin.info/www/bejne/jZuylKicobtpq68 source code: github.com/githubuser1983/algorithmic_python_music/blob/main/bezier_curve.ipynb ). The sound-effects are from LMMS as I felt it sounded good in my ears. It is intentionally bad - so to say.

@@musescore1983 Nice, I'll watch that video! (and if the mix is a conscious "lo-fi" choice, then it's fine)

Thanks!

hahahaha I guess so, that it was not difficult.

thanks

thanks for your comment

Thanks

donkishot

thank you for yor nice comment and words.

Thank you for your comment. I use the k-nearest neighbor algorithm with some scripts in python: github.com/githubuser1983/knn_music/blob/main/knn_music.py The method / algorithm to generate the music is described here: orges-leka.de/knn-music/Measuring_note_similarity_with_positive_definite_kernels.pdf I am experimenting with different algorithmic techniques but my background is more in mathematics not in music. May I ask what your composition style is, more algorithmic or do you compose by hand so to say? Kind regards form Germany.

@@musescore1983 Kind regards from the US although I grew up in Germany ;-) Some of my background can be found here: www.dirktomandl.com/ For my day job I am a scientist (Cheminformatics, AI/ML) in drug discovery. As a hobby composer I focus on classical music in a broad sense. I have used both algorithmic approaches as well as traditional techniques. For example, this playlist (kzbin.info/aero/PLbG6g-0Kt-BdgBGG9DjJ95cy-DbZ3-Wyp) includes piano pieces where I used AI as inspiration and idea generator. The actual compositions were created by me. I consider this approach a true man-machine collaboration. The playlist also includes a video where I explain my process. Recently, I released a piano concerto (kzbin.info/www/bejne/moDFfKendq90fc0) which did not use any algorithms whatsoever. I wrote it completely by hand. As an aside, it got recently performed as well as professionally recorded in a studio. I will post it on KZbin once it is finished. I believe your approach could be especially useful on a microscopic level, i.e., writing a few bars of a melody or filling details for a given musical phrase etc. What all algorithmic approaches so far are lacking is a sense of overall structure. Classical music often uses the so-called "sonata form". Pop music usually uses a verse/chorus form (ABAB) plus intro, outro, and bridges. In order to write "good" music an algorithm needs to understand the overall structure of its composition. For example, it needs to know if it currently is writing the verse or the chorus or a bridge in a song. Otherwise it will be just a sequence of random notes that don't connect with human listeners. "Good" music balances familiar with new: The main melody or theme often gets repeated or varied several times. Then something new happens and then it switches back to the familiar melody. In other words, "bad" music either repeats the same musical fragment over and over again or every note is new without any recognizable pattern. From an algorithmic perspective, I believe the most promising approaches are transformers that are capable in learning long-range patterns in music. Musenet from OpenAI is a good example and perhaps the best algorithmic composer that exists. Your approach could potentially postprocess and improve the music that comes out of transformers. But, again, the biggest gap is the lack of structure in all algorithmic composers.

Thank you very much! :-)

thanks!

thanks! 🙂

thanks for your comment and support!

What is this comment suppose to mean?

Thank you for your comment. The source code is available, so it should count as algorithmic composition. 🙂

You migh also enjoy a shorter piece: kzbin.info/www/bejne/qn7Lo3aYoLOhepY

Dear Greg, If you have contact to someone who would like to perform this live, you are welcome to share the score above the link. It would be nice, if you like, to keep me informed, as I am also courious how it would be performed live.

@@musescore1983 If I come across a group who I think would be interested in performing this piece I'll let you know.

@@GregHarradineComposer Thanks very much! :-)

Thanks Stefaan, The circles are the points for which the Voronoi cells are computed.

Thanks Stefaan, Yes, midi is a rich source for music visualisation.

I have not tried it with python2.7 and on mac. But you have to probably install some packages with pip such as scikit-learn, numpy. Furthermore you have to install with apt-get the sagemath and opencv for python. The shell-scripts with "example" in the name, should show how to use. Otherwise you can dm me: [email protected]

Thanks!

danke!

Thanks, my friend! ;-)

Hi and thanks for your comment. I wrote the script in python, with an algorithm I have developed to chose the arpeggios one after the other and the volume + durations of the notes. Then I put this score with different instruments in lmms by hand. May I ask what your background is, that you ask about python?

This other piece was done with the same method: kzbin.info/www/bejne/q5DZanh4ac-EY7M

@@musescore1983 I am a music nerd and am learning Python. I'm interested in algorithmic composition.

@@BrianHagertyLawyer If you are interested in the scripts, here is some similar script: github.com/githubuser1983/nice_sonification_of_oeis_sequences How it works, is described here: orges-leka.de/knn-music/Measuring_note_similarity_with_positive_definite_kernels.pdf

The sequence of the digits in base 10 determine how many nearest neighbor notes get appended in each step, after having defined a similarity of consonance between notes: orges-leka.de/knn-music/Measuring_note_similarity_with_positive_definite_kernels.pdf

@@musescore1983Thanks so much for that paper i really liked it. Still dont know exactly though, if the algorithm generates chords or not...Id say it does

@@wurnotantmlb Thank you for your nice words. You are welcome. I am working to include chords in the algorithm, which is not very easy because you have to compare multiple notes with one note etc. but I think I have a solution to this, which I can implement in the near future. As for the sounding: Since simple ratios as in chords are prefered by the algorithm, this might sound like a chord, I guess.

@@musescore1983 sorry about my terrible math understanding. I guess its all derived from the digits of pi. Maybe arranged a little, that sort of thing...thanks!!

@@musescore1983 I must apologize, I didnt listened to the piano very carefully the first time and I assumed it had chords but it doesnt...im really sorry!. But it was a good question either way. Im so much into pure data and couldnt be more interested about all this. You have earned a new loyal subscriber here!! Id really wish to know which software are you using if its not an inconvenience or anything. Thanks again!!

thank you for your kind words.

Dear Vladymyr, Thank you for your kind words. Greetings from Germany.

Thanks Stefaan for your feedback and for listening.