Make a song

AI music generation platforms are enabling song creation through a first-principles approach that bypasses traditional music theory. Suno's core technology, for example, models music as a **raw sound wave** sampled at 48,000 times per second, rather than feeding its models concepts like the 12 tones of Western music [12, 19, 23]. This allows the AI to generate novel sounds and blend genres in ways that a more constrained, theory-based system could not [3, 23]. For lyrics, the platform utilizes Large Language Models (LLMs) based on user prompts [1, 20]. Early product decisions prioritized the creation of complete, three-minute lyrical songs over achieving perfect audio fidelity or focusing on instrumental tracks, a contrarian bet that proved more engaging for users [6, 21, 22]. The models currently excel at more formulaic genres like country and pop .

This approach positions AI music generation not as a professional tool but as a new category of "creative entertainment," where the primary value is the user's enjoyment from the act of creation itself [3, 29]. This strategy reframes the product to compete for user time against gaming and social media rather than professional music software, vastly expanding the addressable market . The success of this positioning is reflected in exceptionally high user engagement; on a typical day, **90% of Suno's users** actively create music on the platform [3, 13, 16]. The company's future roadmap aims to deepen this engagement by making music creation a more social experience, with planned features for collaborative creation, sharing song templates, and incorporating users' own voices into the generated tracks [4, 10].

The cultural and commercial impact of this technology is already significant, validating AI's potential to create new pathways in the creative industries . Songs created entirely on Suno's platform have landed on official music charts [2, 5, 9], and some users have secured record deals as a result of their AI-generated work [3, 27]. A partnership with Warner Music Group further signals industry validation and opens possibilities for new products that allow fans to interact with their favorite artists [3, 7, 15]. The technology has also produced viral content, such as a song about Puerto Rico created by a comedian from his travel notes .

Suno's CEO argues that a sustainable competitive moat in creative AI is built on proprietary preference data and rapid research cycles, not on model size or compute scale, which contrasts with the prevailing logic in the LLM space [3, 17, 26, 30]. While acknowledging that larger companies could eventually catch up on model quality, the focus is on user experience and iterating quickly based on user preferences [3, 25]. This suggests that for creative AI applications like music, the key challenge is less about scaling technical capacity and more about solving the research and data problems associated with subjective artistic taste .