By Louise Blas - Loading & Prospective Watch in Plaine Images
Hard hard to follow abundant news around AI, as the outings of products and other "revolutionary" ads abound ... Producers, artists, composers, labels, Data Analysts, Music Tech service developers: AI affects all trades and interferes in artistic creation and operational processes to shake the industry. Water & Music has just given a so -called webinar to take stock.
First an observation: the market becomes mature, with 25% of music creators who have already used AI tools (the majority to make separations from Stems or Mastering, rather than creating a full piece). Industry pros are 40% to use AI in processing their data. In terms of investments, however, there is a substantial slowdown in tickets: the fault of current legal proceedings and the legal limitations that persist.
Three ia trends of music tech
- Faced with the rarity and cost of "real" data, AI models are training more and more on synthetic data. So goodbye license costs and negotiations with rights holders. Nourishing the beast with its own outputs seems at first glance a hazardous idea when you now know that generation after generation, AI degenerates . However, the researchers explore this track for the generation of plugins , instrumental variations or the tagging of data . A hybrid approach mixing "real" and synthetic data seems to be the most optimal to train models. Note that the audio quality generated by the AI is still insufficient to allow "self-born".
- The fine control of the models is progressing, and the user can more and more guide and modify the OUTPUTS of the AI (in particular by modifying a rendering by prompt iteration, without having to regenerate zero with a new more successful prompt). This increasingly advanced granularity erodes according to Water & Music the traditional advantage of composers and producers, with an increased place for the artistic direction, to the detriment of technical expertise. A trend similar to Rick Rubin's magic recipe, which has already spoken on this subject .
- Finally, the detection of content IA marks a decisive turning point, in direct connection with the protection of copyright: streaming platforms track artificial content, services offer certifications for data training models, musical deepfake deepfake detectors content watermarking
In summary, power relations between a multitude of actors are built and it remains difficult to predict which will hold the controllers of AI in the near future. And Water & Music to conclude with a vision that will not fail to react: artists will probably have to adopt a more flexible point of view with regard to intellectual property, in order to improve their resonance, whether artistically or operationally.
