Skip to content
AudioStellar, a trip to the audio galaxies An open source music tool for exploring sounds using visualization

AudioStellar, a trip to the audio galaxies An open source music tool for exploring sounds using visualization

imagine a Free digital music instruments and software,. Now add that it uses techniques of Artificial intelligence and which allows to try out the audio content. With that add a visual interface which delivers a huge, colorful constellation in which different sectors are identified by the wave characters of each sound. Incorporate the possibility of using protocols of communicate with other devicesincluding your cell phone. Finally, imagine it as a local development through a public university. That is SoundStargy.

This is a research project brought forward by Leandro Garber within the framework of MUNTREF Center for Art and Science, of Tres de Febrero National University. They are working with him Tomás Ciccola and Juan Cruz Amusateguiresearch students and collaborators in the development of this project focused on the search for new wood, textures and compositional strategies using cutting-edge technologies.

SoundStargy processes a folder with sounds and generates a map placing each sound as a point in the 2D space. Adjacent points correspond to similar sounds from a spectral point of view (that is, to similar waves), while distant points represent different sounds.

“It simply came to our notice then explore audio libraries or field recordingswith the help of artificial intelligence which shows the existing structure in the relationships between the audios, and on the other hand play the audios through new logic and criteria that cannot be achieved by using traditional techniques ”, says Leandro.

Step i Electronic Arts and specialized in Machine LearningGarber explains that two aspects can be identified in the project: “An artistic, performance – oriented, expressive search for new sounds through non – traditional sound design and synthesis strategies; and another technological / scientific one, which involved new techniques data analysis through artificial neural networksinteracts with the artistic side ”.

Garber reviews the three general modes of AudioStellar, after more than a year and a half of commitment:

  1. 1. Scanning modewhich allows you to analyze the map and listen to the suggested relationships.
  2. 2. particle methodwhich gives the possibility to issue mobile and configurable autonomous agents, which are capable of reproducing sound from their movement on the audio map.
  3. 3. sequence modeused to draw constellations between the points located on the plane and reproduce them sequentially, using the distance as a space between the sounds, generating rhythms.

there are all methods compatible with MIDI and OSC, the most common hardware communication protocols, which means that various physical controllers, such as a mobile device or tablet, can be connected and interact directly with AudioStellar. “These first general methods are within them hundreds of different possible configurations and are designed to be very flexible, “says Garber.

Leandro Garber is an electronic artist and machine learning researcher Photo: Cecilia Salas

Free license for sound processing

The artistic resource and many of those can be followed in the AudioStellar Facebook Group. Here, users post their tests, like the musician and the producer Agustin Spinetto (former member of rap duo El Orgullo de Mama), originally from Tokyo uses AudioStellar as an interface to control analogue syntheses. In fact, it’s a a central feature in his thesis for the Master of Musical Creativity and the Environmentfrom the University of the Arts in the Japanese capital.

In that same group, Garber put up a video of collaborator Tomás Ciccola with Antonella Casanovashowing one of the most interesting ways to use AudioStellar: premiere record and then cut it into short sections. Once the folder of these small audios is located in AudioStellar, “we can hear how the AI ​​algorithm is able to separate vowels, consonants, breath, empty spaces and micro-noise“.

In line with free software values, AudioStellar is open source and license freemaking it an interdisciplinary research platform anyone can download, study, modify and distribute as they wish. “The project attracted people from a variety of fields: sound art, music, musicology, programming, sound engineering, data science, so keep this license and my main interest is not to focus on commercial proprietary software“, Garber dice.

AudioStellar competed and was accepted at the international conference New Interfaces for Musical Performance (NIME) of 2019. It was also part of the sound art symposium UNTREF sound world and as a scientific document in the Machine Learning Summer School (MLSS) 2018 at Di Tella Institute.

Currently it is open Call looking for an audio artist to join the AudioStellar project, co-design, experiment with the software and document these experiences. “It simply came to our notice then study and identify gesture-expressive veins of the projectilegenerate specific instruments from the general methods, present sound performances used by the device and that dialogue with other contemporary instruments “, Garber concludes.

Start investing in Stellar and get Free Bitcoin when you buy or sell 100$ or more when you register in Coinbase