Google Launches AI Music Creation

- The technology behind this launch, Lyria 3, is part of Google's long-running research in AI music generation, which began as a research project called Magenta in 2016. Previous versions of Lyria were tested with musicians and YouTube creators before this wider public release in the Gemini app. - Generated tracks are 30 seconds long and are embedded with SynthID, a type of inaudible digital watermark that allows Google to identify the audio as AI-generated. The model can create music from text, images, and videos, and it generates its own lyrics and cover art. - The launch enters a competitive landscape of AI music generation, with established players like Suno and Udio being recognized for producing high-quality, full-length tracks. Other notable AI music tools include OpenAI's MuseNet, AIVA, and Amper Music. - Under current U.S. Copyright Office guidance, music generated entirely by AI is considered to be in the public domain and cannot be copyrighted. However, works that involve a significant amount of human creative input alongside AI assistance may be eligible for copyright protection. - The release has been met with mixed reactions from the music community; some artists express concern that AI could devalue human creativity and lead to a flood of low-quality content. Others view it as a potential collaborative tool that can assist in the creative process. - In addition to being in the Gemini app, the Lyria 3 model is also integrated into YouTube Shorts through a feature called Dream Track, allowing creators to generate custom soundtracks for their videos. - Google has stated it trained Lyria 3 on music that it has the right to use through its terms of service and partner agreements. This is a sensitive issue, as the practice of training AI models on large datasets of existing, often copyrighted, music is a point of contention in the industry.