Oct 29, 2019

Connections 2: CECIA

In our new article series we present projects and collaborations that connect to Ulysses Network projects or members. The second article introduces CECIA project through an interview with Kosmas Giannoutakis and comments from participating composers. You can find Connections 1 article about reConvert here.

Do you have a ULYSSES connected project you would like to tell about? Contact the community manager community.ulysses(at)gmail.com!


Cecia diagramDiagram of the CECIA Project

What the project is about?

Collaborative Electroacoustic Composition with Intelligent Agents (CECIA) is an innovative art project that integrates the creative agency of 5 composers and Machine Learning algorithms in order to produce a coherent composition of electroacoustic music. Organized by ZKM within the framework of the »Interfaces« project, with the support of the Creative Europe program of the European Union, the project’s goal is to explore collaborative music creation, by connecting composers and Intelligent Agents through a cloud platform. With features such as constant web presence, transparent creative process and well-documented collaborative activity it aims to produce content with high artistic and research impact for musicians, composers and music enthusiasts.

In October the project is halfway through. What has happened until now?

The project started with an open call for composers, in which composers of electroacoustic music were invited to submit their CV, two recent works and a motivation statement. We received around 100 applications and we end up forming a diverse group of composers, consisted of Panayiotis Kokoras, Mariam Gviniashvili, Juan Carlos Vasquez, Martyna Kosecka and Erik Nyström.

During a residency in July at ZKM, my colleague Artemi Maria Gioti developed the Machine Learning algorithms while I was responsible for the development of the cloud platform. Together we designed a rigorous framework that makes possible a systematic collaborative workflow. In our model the ML algorithms facilitate the creation of new material while the composers exchange multi-modal information following democratic procedures.

Since the beginning of the collaborative activity with the composers in August, we have created a pool with 25 original recordings of total duration of 38 minutes, 35 composed phrases of total duration of 20 minutes and 70 algorithmically generated phrases of total duration of 1 hour and 16 minutes. All this material has been repeatedly evaluated and processed, leading to the crystallization of a draft electroacoustic composition of total duration of 4 minutes and 20 seconds (https://cecia.net/wp-content/uploads/2019/10/underworld.mp3). You can find all the material, read the discussion that led to that output and follow the further development of the project at https://cecia.net.

What will happen during the rest of the project?

Our goal until the end of November, is to reach 60 composed phrases, 120 generated phrases and achieve a length of approximately 8 minutes for the draft composition. After that we will start a revision phase where the composers will apply changes throughout the draft composition. The final step will be the finding of an appropriate title that reflects the nature of the finalized composition.

What have you learned during the project, and has it changed somehow the aims or methods?

During the conceptualization period of the project, we tended to consider the Machine Learning algorithms as “virtual composers” that contribute with the human composers to the creative process on equal terms. In our collaborative model, the exchange of all the material takes place anonymously so we thought that the project could be sort of a “Turing test” for algorithmic composition, that asses the applicability of such algorithms in the compositional process. Soon we realize that the music created by A.I. algorithms lack the multifaceted qualities that characterize the music composed by professional composers and it would not be reasonable to bring them in a competitive relation. So we changed the plan and we gave the role of the “assistant” to the ML algorithms that learn some features from the aesthetic choices of the composers and propose to them novel combinations of the available material. This approach have been so far well received by some composers, who have borrowed excerpts from the generated material and incorporated them into their composed phrases.

Will the project be presented somewhere after it is finished?

The final composition will be presented on 14th of December at ZKM Cube as part of a two day festival on telematic art projects and networked music. At the same day I will make a lecture presentation of the project explaining the collaborative model and the ML algorithms in detail, and addressing the research questions of the project. After that we will look for further opportunities of presenting the project in various academic and artistic contexts.

Is there a continuation for the project in form or another?

The next collaborative composition project I am organizing, will take place from November 2019 until May 2020, with a focus on the composition of instrumental music for chamber ensemble. It is funded by University of Music and Performing Arts Graz (KUG) and supported by the Institute for Electronic Music and Acoustics (IEM) under the supervision of professor Robert Höldrich. The composers participating in this new project are Lula Romero, Márcio Steuernagel, Karin Wetzel, Stylianos Dimou and Davor Branimir Vincze. The final composition is going to be performed in May in Graz by the Schallfeld ensemble.

Comments on the project by the composers

"The CECIA project present unique characteristics among the current efforts of pair AI with music composition. Since the conception, the project did the bold step of commissioning a group of electroacoustic composers to work on a double-blind peer-reviewed piece in 30 second phases, a process that represented an excellent pool of data for the Machine Learning algorithms to learn the intricacies of acousmatic composition. As a result, the algorithms have done a remarkable job so far, providing assistance to the composition process and improving consistently towards a near human-like imagination. I have worked before with a network of intelligent agents in the past, but never paired with a network of human composers in such an organic and impactful way. It’s been a significant process both individually and collectively, and I don’t doubt that the final results will be artistically commendable." - Juan Carlos Vasquez


Juan carlos VasquezComposer Juan Carlos Vasquez


“CECIA gave me an opportunity to take part in a well-organised collaborative composition. It was a unique experience to further develop other artists' creations right in the same workflow and get insights of their way of composing. As the pool of our musical phrases got larger and more diverse we could witness an interesting evolution of the AI agent's creative output.” - Mariam Gviniashvili

“I started to work in CECIA project as one of five collaborating with each other composer fellows under the guidance of two other artists. After completing more than two months of collaboration, I know now that this project is highly innovative, very demanding at times, and its weekly results still surprise me as an artist. We compose in a routine way, thirty seconds every week, we compare the results, listen to machine generated suggestions, decide on what to follow for later in musical construction, and try to spice it the way our imagination tells us. May sound ordinary, but for the first time I create music that in not one hundred percent under my control, or so to say - its result or continuation is a matter of a common taste, careful ear while listening to everybody's sounds, and democratic decisions of a group. Or - maybe it's always decided by the algorithm we're not aware of?” - Martyna Kosecka


Mariam GviniashviliComposer Mariam Gviniashvili


Martyna KoseckaComposer Martyna Kosecka. Photo by Tiange Zhou.

“This is a very interesting and quite challenging project, which raises both technological and aesthetic questions. When AI agents learn from a diverse aesthetic input from human composers, what will they perceive as a model for electroacoustic composition? Are we teaching them artistry or habits? Will they also introduce they own aesthetic into the blend?” - Erik Nyström

“It has been great experience participating in CECIA, a Collaborative Electroacoustic Composition with Intelligent Agents. It is fascinating to see how machine learning generated composition materials, help us to understand better our compositional choices, and enhance the creative process.” - Panayiotis Kokoras


Erik NyströmComposer Erik Nyström


Panayiotis KokorasComposer Panayiotis Kokoras

Associated members

Panayiotis KOKORAS
Composer, Conductor, ...