THAT VOICE IS BEAUTIFUL. BUT WHO IS SINGING?
Virtual singers, AI-generated music and credits to explore. A short guide to discovering how a song comes together.
The chorus wins you over. The cover shows a singer and name; everything looks familiar. Then you discover a synthetic voice. Character, sound and lyrics may have different origins, making the credits worth exploring.
From text to song
Systems such as Google DeepMind’s Lyria 3, introduced in February 2026, can generate music and lyrics from a request. Users may then edit words, assembly or performance. Saying only “made with AI” does not describe all those steps.
Ask which parts were generated and which written, played or adjusted by people. Two tracks made with one tool can arise in very different ways.
An invented character or an imitated voice?
A virtual singer does not necessarily copy a real artist. Recognisably reproducing someone’s voice or face is another matter, addressed by the US Copyright Office in a 2024 report.
Before sharing a supposed collaboration between stars, check official channels. An “X sings Y” title somewhere online may describe a simulation: resemblance does not mean participation.
What about rights?
In a 2025 report, the same office distinguishes human contribution from generated material. AI use does not remove protection from human-created parts, but its analysis says a written prompt alone is insufficient. This concerns US law and case-by-case assessment.
Your ears cannot tell you everything
Google’s SynthID embeds an inaudible watermark in audio from its systems, detectable with dedicated tools. You cannot hear it in a chorus, and it does not cover every track online.
Next time, look for who wrote the lyrics, AI’s role and who releases the track. Then enjoy the music! Loving a song and wanting to understand its makers can go together.
Source: Google DeepMind ↗
