Suno Launches Speech Model for End-to-End Voice and Music Generation

US-based AI music platform Suno announced the launch of its Speech audio model on October 1. The company calls it the industry's first end-to-end closed-source audio model that can generate voice and music together as a single coherent track. Unlike traditional workflows that first do text-to-speech and then add background music, this model generates a complete blend of spoken voice and original background music in one pass.

After a month of testing with a small group of users, the feature is now open to everyone in beta and built directly into Suno. Users just need to input an idea, a poem, or some text, then describe the desired voice tone and music style.

Suno also noted current beta limitations: occasional accent instability, such as a British accent drifting to Australian and back mid-track; and intonation and emotional pauses that can sometimes feel overly dramatic, causing slow pacing in sentence breaks.