First prediction - part 4

A pure data designed algorithm (part 1) generating about 10 hours of midi data using five different configurations (conditional statement alterations in the same key & pitch) to produce a dichotomy in the 10 sperate hour long performances. From these 10 hours of midi performances around 150 midi files between 4-35kb in size have been edited out, constructing a synthetic dataset almost 5mb in size. At this point, all of the original midi song files have been broken up into individual instrument files; drums, bass, chords & melody. Each instrument assigned to a specific instrument dataset. In total there are 4 datasets of 150 midi files. I convert all the midi files with Music21 into text and then back to midi after prediction. Breaking up the instruments into separate midi files allows for moving the tracks around within datasets for different LLM models (GPT&LSTM) and with the ongoing arraigning of the current score in development.

The video in this post is a unique one. This is the first prediction made from the original island midi score trained on the synthetic dataset.

TRAINING STATS

number of training files = 150 - batch size = 30 files - number of iterations (number of training files/batch) = 5 - one epoch = 5 iterations

The bass guitar heard in this video is the first instrument predicted for a single epoch. For this audio, this first predicted track has been placed it back into the original midi score. Comparing the bass guitar in the island video of my previous post (part 3) to the bass guitar in this video (part 4) demonstrates a first step using the LLM to predict a new midi score for this single instrument in the group.

Note the interesting artifacts of the bass guitar performance. While these notes are in key, they are reminiscent of stray(bad) midi notes often produced in live midi performances. Velocity and note durations can be a prediction issue. The predicted outcome will change even more so with increased epochs. Ai prediction is a very cyclic procedure. When auditioning and recording these midi outcomes, little attention is given to the audio. Please excuse the lower quality instrument sounds used to demonstrate in this video.

Also to note, are the different drum performances heard in the two videos. Drum pattern prediction differs from note prediction and will be explained in a future post.

initiating a synthetic dataset - part 3

With a little modifying of the conditional statements & performance auditioning of the pure data patch presented in part 1 of my posts, the results have produced this song (groove): island. There was no preconceived idea for this song. The song comes about from building the pure data patch and listening to the results. All of my algorithm patches are structured in a 4-piece instrument band concept. Drums, bass, left hand chords & right-hand melodies. I'm structuring my algorithms for improvised pop genre song types. Pure data song design patches can be structured for any genre. In pure data patch design, I strive on making the generated instrument performance sound like a band in key. Achieving results that sound like music from different genres takes some tooling. Creating an algorithm that generates a specific type of song is only limited by your understanding of MAX, pure data or other midi generating system and music.

The drums are set to fixed scale (can be any key) midi percussion zone configuration. Which helps out greatly when drum pattern predicting. Structuring the inbound generated midi drum notes to a fixed scale/ zone will make it easy to set up an outbound predicted fixed scale/ zone from the AI model. This way every predicted midi note will be assigned to a percussion instrument in your midi rig. Even if it's scrambled you might hear what you are looking for. Drum pattern prediction can produce some zany results. The bass & right-hand piano parts are broken up into two separate harmonious instrument sequences. The chords are the total number of triads (three note combinations (around 20)) are predetermined from the scale (F#). These chord combinations are selected when triggered from the patch during performance, with a harmonic relation to the notes chosen in the bass & melody sequences. Always room for design improvement when it comes to harmony. A closer look at the harmony of the left & right hand of a piano is examined in my first music video post, "day".

For the purpose of presenting how I use AI to create original (no copyright issues) music in this series of posts, "Island" represents the song that will be used to initiate a synthetic dataset for AI prediction. The creative initiative if you will for the dataset's theme. Representing a point in the song composing process where the artist decides whether or not to start building a dataset based on the algorithm's design performance. Or continue on the algorithm in pure data to achieve something different.

Ethical Ai - call to create synthetic datasets

Ethical Ai - a call to create original "synthetic" datasets.

With regards to music prediction, what is ethical Ai?

The issue lies in the datasets utilized. The ethical considerations surrounding the technology underpinning large language models (LLMs) are often overstated. LLMs derive their predictive capabilities related to syntax, semantics, and ontologies embedded within human-generated corpora; however, inheriting inaccuracies and biases present in the training data. The nature of the learning data within a given dataset ultimately determines whether the outputs produced by the LLM are deemed ethical. If a organization develops an AI system and generates modified versions of existing musical works for personal enjoyment, this practice is generally regarded as ethical. Conversely, if the objective is to distribute or profit from these alternative renditions of established music, such actions would be considered unethical.

A studio might develop two moral AI systems. First, a dataset of all the music the artist has produced over the years. Based on earlier compositions, this technique could be used to forecast new music ideas or compositions. An LLM system that uses a dataset made out of created input data is the endeavor. I'm going to demonstrate a system that can be applied to the quest for new music. As AI prediction advances and the original dataset and predictions are re-incorporated into the dataset, the second idea ultimately becomes the first.

There are many ways to generate MIDI input data for AI datasets, including sequencers, drum machines, MAX, Pure Data, Audiomulch, noise, voltages, and scientific data. The potential for discovering new music genres is limitless when exploring the predicted MIDI outputs derived from these original sources. It is clear that large language models (LLMs) can play a significant role in music exploration and can facilitate the creation of new compositions from original datasets. The music industry would benefit from more artists developing their own AI models and utilizing unique datasets to advance this emerging genre. I employ Pure Data to generate my MIDI input, focusing on structuring algorithms that produce MIDI data sculpted to a genre type and in a given musical key(A#b). I advocate the importance of artists building their own datasets and moving away from a reliance on historical and commercial data to produce music.

Technologies establish cyclical procedures for our adherence. Engaging in the iterative refinement of a process is essential to achieving the desired outcome. Is there work to be accomplished? Indeed, determining the specific type of input data required for your dataset and devising methods for generating that data is a time-intensive endeavor. If the objective is to create something original, like band practice, it becomes a labor of love.

dataset procedures that are suggested online

pure data polyrhythmic metronome - part 2

With a little modifying of the conditional statements & performance auditioning of the pure data patch presented in part 1 of my posts, the results have produced this song (groove): island. There was no preconceived idea for this song. The song comes about from building the pure data patch and listening to the results. All of my algorithm patches are structured in a 4-piece instrument band concept. Drums, bass, left hand chords & right-hand melodies. I'm structuring my algorithms for improvised pop genre song types. Pure data song design patches can be structured for any genre. In pure data patch design, I strive on making the generated instrument performance sound like a band in key. Achieving results that sound like music from different genres takes some tooling. Creating an algorithm that generates a specific type of song is only limited by your understanding of MAX, pure data or other midi generating system and music.

The drums are set to fixed scale (can be any scale) midi percussion zone configuration. Which helps out greatly when drum pattern predicting. Structuring the inbound generated midi drum notes to a fixed scale/ octave will make it easy to set up an outbound predicted fixed scale/ octave from the AI model. This way every predicted midi note will be assigned to a percussion instrument in your midi rig. Even if it's scrambled you might hear what you are looking for. Drum pattern prediction can produce some zany results. The bass & right-hand piano parts are broken up into two separate harmonious instrument sequences. The chords are the total number of triads (three note "triad" combinations (around 20)) are predetermined from the scale (F#). These chord combinations are selected when triggered from the patch during performance, with a harmonic relation to the notes chosen in the bass & melody sequences. Always room for design improvement when it comes to harmony. A closer look at the harmony of the left & right hand of a piano is examined in my first music video post, "day".

For the purpose of presenting how I use AI to create original (no copyright issues) music in this series of posts, "Island" represents the song that will be used to start a synthetic dataset for AI prediction. The creative initiative if you will for the dataset's theme. Representing a point in the song composing process where the artist decides whether or not to start building a dataset based on the algorithm's design performance. Or continue on the algorithm in pure data to achieve something different.

Creating midi input datasets with pure data - part 1

One of the drivers of my KicKRaTT (linktree in profile) project is examining the prediction choices AI makes from generated synthetic midi input data. To predict an original song that can be genre identified over that which would be deemed experimental. ARBOREAL was constructed from the predictions made by GPT & LSTM models trained on datasets created by two algorithms generating midi input data in the key of C. KicKRaTT's next AI composition will be composed by algorithms generating different genre type midi input in the key of A major. There are many ways that you can generate midi input data for AI synthetic dataset constructions. The algorithms that I use to generate the midi input for the AI training datasets are constructed in pure data. This video demonstrates a patch that has pitch adjustment, adjustable conditional statements & will can generate midi for any length of time. Generating synthetic midi for AI datasets could also be done in Max, audiomulch, software synthesizer modules & vsts. There are numerous ways to generate midi or quantize other sources to midi. Using unorthodox sources for generating original midi data opens up many new genres of music available to predicting.

If you going to teach your computer to compose, use original data, & it will predict with characteristic not plagiaristic. Training your LLMs with datasets composed of known works of music ends in recycled music.

This midi generation video is performed using the MS GS Wavetable available on all Windows desktops. At this point in this process, I am only focused on sculpting the algorithms design to generate midi in a specific genre type or to achieve a desired groove.

The process has been started to compose an entry for the International Ai Song Writing contest of 2025. From here forward, to the entry submission data, the blog posts will present in steps, the compositional process. Posts that have started on LinkedIn will be duplicated here on Blogger & on the journal. The contest entry will demonstrate an ethical use of Ai prediction across three genres of music with a change of key. The music and the video will be conceptually integrated. Aspects of the video process will not be posted until after the contest.