Welcome to the KicKRaTT web on electronic music composition. Generating midi notes using algorithms, commercial devices, noise and conversion data. This journal presents the different ways I generate midi music, the compositions & online activity of KicKRaTT. Structuring algorithms to generate musically harmonized composition. Developing a process to compose intelligent music & bring structure to chaos RANDOM / CONDITIONAL music generation. At present, I am generating MIDI with algorithms structured in the pure data (pd) environment. Aka, open-source MAX. Generating MIDI in this way produces solos for individual instruments & multi-instrument synthetic bands, infinite improvise and datasets. I use the generated midi to orchestrate the audio components in my studio, construct synthetic datasets for large language model (LLM) training, improvisational live jam sessions and genre-specific music generation. I hope to cover numerous midi generating schemes that produce humanistic instrumentals, explore LLM technology to predict composition variations and my continued work on my "evolving musical algorithm". An algorithm that generates music and changes its conditional statements in real time to improve the musicality of the composition output.
In the audio department it is all about guitar & bass sounds. Part two, titled KaOzBrD will be a dramatic shift towards a guitar centric improvisational band. I have made serious progress in the guitar & bass sound that will change the KicKRaTT sound completely. One might even say the genre! Stay tuned!
This Welcome Page serves as a gateway to the different online presentations related to the KicKRaTT music project and the links found on the LinkTree page. If you have arrived here via platforms such as Linktree, SoundCloud, SoundClick, YouTube, or Vimeo you will find all the music associated with this project uploaded to the KicKRaTT and KaOzBrD SoundCloud, SoundClick & ReverbNation accounts. At these accounts can be heard all of the music discussed on this journal, featuring both new works and previous relevant compositions. On YouTube, Vimeo, and Odysee, you will find videos for the generated, predicted & performed music at various developmental stages. KickRaTT music can be purchased on the SoundClick and Bandcamp.
This Google Blogger is the straight-forward version of the KicKRaTT Journal.
The current project involves generating midi from algorithms structured in the pure data (pd) programming environment. Pure Data an alternative to MAX MSP can be downloaded online. The focus is on designing algorithms that generate midi through random, conditional and expression-based mathematics. The structured algorithms create endless midi music. The project's objectives are to craft a musical style into the algorithm's design that evolves during the generative process. To generate a definable form of music that easily fits into mainstream music genre categories. To produce computer-generated compositions that exhibit a distinctly human-like quality. I will be documenting this journey here on the Dreamwidth journal (covering the process and conversations) and on Google blogger (streamlined) to provide insight, conversation and variety.
Thank you for enjoying the music.
KicKRaTT has taken part in the International AI songwriting competitions of 2024 & 2025. The 2024 AI song submission "ARBOREAL" was composed with pure data algorithms generating enough midi input to compose synthetic learning datasets for the large language models (LLMs) used to predict. The pure data structure essentially generating its own learning dataset and then generates the midi input for the LLMs to predict off of. The midi music compositional process blooms in a self-generative way in this procedure. It's a unique procedure and has been recognized as a first in auto-generated music composition by the Ai community. The process was elaborated on for the 2025 Ai completion entry SOLARIUM. Generating learning datasets with algorithms & various commercial midi devices, Utilizing LLMs, specifically GPT and LSTM models, trained on the generated midi input datasets, composing a final score through predicted variations. The goal was for an algorithm to generate the LLM training datasets and for the LLMs to predict a composition based on the continue generated midi input. The algorithm or device that created the dataset also created the midi input that the LLM used to predict. While minor adjustments to the conditional statements of the algorithm during the process, the statement is true that the same algorithm or device that created the training dataset is the same that wrote the song. An original score from a fresh generated source of data, avoiding the use of commercial or historical audio samples or MIDI files to train the AI models.
Constructing the AI workstation for the GitHub LLMs to reside, developing the pure data structures to generate the MIDI input datasets, and creating an AI training program that evolved the algorithm composition through predicted variations and managed datasets was the endeavor. The process and competition participations are presented and discussed in detail on these journal entries, Google Blogger, and YouTube videos. This developed process has been presented, examined & judged in the two Ai competitions of 2024 & 2025.
If it is the competition tracks ARBOREAL or SOLARIUM that has brought your interest here, then proceed to one of these links below or the YouTube & Soundcloud KaOzBrD icons at the bottom of this entry. The song ARBOREAL and the Ai song contest experience is pretty well documented here.
A brief reporting on the International Ai Song Contest of 2024.
Ai Song Contest 2024 ARBOREAL Process Document.
A brief reporting on the International Ai Song Contest of 2025.
Ai Song Contest 2025 ARBOREAL Process Document.
Thoughts about constructing Ai datasets.
* KicKRaTT; MUSIC, ALGORITHMS, DOCUMENTS, GRAPHICS & LOGOS: ISNI# 0000000514284662 *
No comments:
Post a Comment