Benoit Carré, the former member of French group Lilicub, has spent more than a decade exploring musical creation with artificial intelligence prototypes developed alongside researchers. After testing Suno, a generative music platform, he concluded that its model is technically better than he is at producing complete songs in a wide range of styles.
Suno says users can create any song they imagine, including prompts built around unusual subjects and genres. The platform generates more than 7 million songs a day and reports annual revenue of $300 million, with a valuation of $5.4 billion. None of those songs has reached the top of the Billboard Hot 100, although AI-linked artists such as American singer Xania Monet have accumulated millions of streams.
Carré describes the result as both impressive and unsettling. Suno can produce songs with vocals that imitate the emotional qualities and conventions of different genres, prompting questions about what continues to distinguish human musicians from machines.
He compared the system’s output with his own experience composing “Voyage en Italie,” the 1996 Lilicub song created with his partner. The initial verse came together quickly, but the chorus felt unbalanced. At the urging of producer Marc Lumbroso, Carré spent several days reworking the song before deciding that a long, complex “anti-chorus” was needed to counter the light, almost childlike verse.
Carré then asked Suno to create a song about travelling in Italy, emphasizing a contrast between verse and chorus. The tool quickly produced gentle retro songs, but he found that they became monotonous after the first 20 seconds. For him, the experiment illustrated the gap between convincing imitation and the deeper choices involved in shaping a song.
Sources: Le Monde Culture.