Suno launches Speech for end-to-end voice-and-music tracks

On October 1, AI music platform Suno announced Speech, which it describes as the industry's first audio model capable of generating speech and music together as a single, continuous track. According to a report published on October 3, the model departs from the traditional workflow in which text-to-speech output is produced first and BGM is stitched in afterwards. Instead, Speech generates everything end to end, and Suno calls it the first closed-source end-to-end audio model able to produce a fully blended track of spoken narration and original BGM in a single pass.

Over the month before the announcement, Suno tested the feature with a small group of users. It is now open to everyone as a public beta, and it ships directly inside Suno. To use it, a person supplies a prompt — an idea, a poem, or a passage of text — and then describes the voice and the musical style they have in mind. Suno says this makes it straightforward to create spoken audio with backing music.

Suno has also acknowledged the rough edges of the current build. The company says the beta occasionally drifts in accent, offering the example of a British accent sliding into Australian partway through and then drifting back. Intonation and emotional pauses, it notes, can also come across as overly dramatic.

Those claims come from Suno itself, as relayed in the reporting, and the reports do not go into further technical detail about the model or say what comes after the beta period. The release is positioned around a single idea: rather than assembling speech and music from separate steps, Suno is offering one model that produces the combined track in one go, and letting anyone try it while the company continues to work on accuracy and delivery.