Suno Launches Speech, Generating Voice and Backing Music in One Pass

AI music service Suno unveiled the beta of “Speech,” a feature that generates a speaking voice and backing music together, on October 1 local time. Put in text or an idea and you get back a single track containing both voice and music. A service built around music has now widened its scope to spoken content such as narration, storytelling and guided meditation.

There are two ways to use it. You can paste in a script you wrote yourself, or simply describe what you want made. Add the voice and music style you have in mind, and Speech generates the speaking voice and fresh backing music all at once.

Suno says it tested Speech with a small group of users for about a month before opening the beta to everyone on the day of the announcement. It is available on web, iOS and Android.

How it differs from the existing Voices

“Voices,” which came first, is a feature for using your own voice in a song. You record your voice or upload an audio file. Speech, by contrast, builds a speaking voice and backing music from text. Where Voices focused on the voice that sings, Speech is about producing content that delivers a message through speech.

Combining voice and music is something other services can do as well. ElevenLabs Studio lets you generate and arrange voice narration and background music inside a single editor. The difference Suno is claiming for Speech is that the speaking voice and the backing music are generated together in one pass. Instead of preparing the two separately and stitching them together, you get a result containing both elements at once.

For anyone setting out to make narration or a guided meditation, the production process changes. All you have to write down is the content you want to convey, the voice, and the mood of the music. Suno cited storytelling and guided meditation as example use cases. The focus is not on output that sings lyrics, but on output that delivers written text as speech.

Will the pronunciation and pauses sound natural?

Suno flagged the beta’s weak points up front. The company said accent characteristics can waver, as when a British accent drifts into an Australian one and then back again, and that pauses in the middle of speech can run excessively long. Even though you can describe the voice and the music style in words, the only way to know how the generated voice reads your script — its pronunciation and its pacing — is to listen for yourself.

With Speech, the range of content you can make in Suno has broadened from songs toward speech. What remains to be seen is how useful the approach of generating voice and backing music in one pass turns out to be for narration and storytelling work. For readers in South Korea, the practical tests will be how natural Korean-language narration sounds, and what the terms and the cost of using it from a Korean account turn out to be.