Youtube recently introduced a number of features for creating new AI-based music and sounds. Having reached an agreement with some artists, YouTube will use artificial intelligence tools to clone the voices of these artists and eventually produce songs. Google also announced the SynthID tool to detect whether artificial intelligence is used in such music content.
The era of inaudible watermarks for AI music
Audio created using Google DeepMind’s AI Lyria model will be watermarked with SynthID, allowing people to determine whether the content was created by AI. In its blog post, DeepMind said that the watermark cannot be detected by the human ear and will not compromise the listening experience. This particular watermark will still be detectable even if the audio track is compressed, sped up, or edited in various ways.Watermarking tools like SynthID are seen as an important protection against some of the harms of generative AI. US President Joe Biden had called for watermarking artificial intelligence-generated content with his signature executive order in recent weeks. According to DeepMind, SynthID’s audio application works by converting a sound wave into a two-dimensional visualization that shows how the frequency spectrum in a sound evolves over time. It is claimed that this approach is unlike anything that exists today. Google announced a very similar feature for images a few months ago.
47417
