Periodic Reporting for period 1 - FlowMachines (Flow Machines)
Berichtszeitraum: 2015-03-01 bis 2016-08-31
This POC project was about exploring the commercialization potential of style imitation music technologies. Based on a number of successful prototypes obtained in the corresponding ERC project (Flow-Machines) we have explored how to monetize such music generation software, in particular in the music synchronization business (i.e. music composed to be synchronized to existing media such as advertisements, movie music or video games). This exploration had two components: 1) the realization of a number of prototypes and 2) the exploration of the legal issues involved.
During the project, we realized the following:
- The legal aspects of music style imitation are extremely complex (at least in a country like France, but also in other European countries), due to the aging copyright retribution system. A report lists all the copyrights issues involved and summarizes them a very big table. However, we considered that it was not worth addressing the issues raised by each copyright type, in a context where AI technologies are evolving very fast, without risking to produce irrelevant conclusions.
- Direct commercial exploitation of AI technologies per se is no longer a profitable route. Most AI technologies are now essentially free (e.g. TensorFlow by Google). Exploiting these technologies as commercial assets is therefore extremely risky in this very competitive environment.
Instead, we have shifted to a model in which we produce, distribute and monetize content using the "traditional means" of the music industry. Based on all our experience, results, and demonstrations, as well as the tremendous success of 2 AI generated songs (Daddy's car, a song in the style of the Beatles produced in the Flow-Machines project with the help of this POC project got over 1 million views on YouTube, and generated considerable media attention), we have managed to convince top executives of a major music label to create a dedicated label, specialized in the production and dissemination of music composed with AI technologies. This is the route we will now pursue.
During the project, we realized the following:
- The legal aspects of music style imitation are extremely complex (at least in a country like France, but also in other European countries), due to the aging copyright retribution system. A report lists all the copyrights issues involved and summarizes them a very big table. However, we considered that it was not worth addressing the issues raised by each copyright type, in a context where AI technologies are evolving very fast, without risking to produce irrelevant conclusions.
- Direct commercial exploitation of AI technologies per se is no longer a profitable route. Most AI technologies are now essentially free (e.g. TensorFlow by Google). Exploiting these technologies as commercial assets is therefore extremely risky in this very competitive environment.
Instead, we have shifted to a model in which we produce, distribute and monetize content using the "traditional means" of the music industry. Based on all our experience, results, and demonstrations, as well as the tremendous success of 2 AI generated songs (Daddy's car, a song in the style of the Beatles produced in the Flow-Machines project with the help of this POC project got over 1 million views on YouTube, and generated considerable media attention), we have managed to convince top executives of a major music label to create a dedicated label, specialized in the production and dissemination of music composed with AI technologies. This is the route we will now pursue.