From text to accompanied voice, Suno adds Speech to its generator

Suno is expanding its audio generator with Speech, a model capable of producing a spoken voice and its accompanying music within a single generation. The feature is arriving in beta after a month of testing with a select group of users.

Joint Generation for Voice and Music

A poem, a story, or a simple voice note can now be transformed into spoken content accompanied by an original soundtrack.

With Speech, Suno does not treat voice and music as two elements to be produced separately. The model generates both within a single audio track in order to match the vocal performance with the musical accompaniment.

Suno presents it as the first audio model designed to jointly generate these two components. This claim comes from the company and is not accompanied by any comparative benchmark in the published information. Describing the Voice and the Musical Universe

The workflow uses an interaction already familiar to users of the platform. A text or an idea serves as a starting point, then a prompt describes the desired voice and the musical style that should accompany it.

Speech can thus produce dramatic readings, meditations, poems, short speeches, or stories, with music generated specifically for the spoken sequence.

The approach differs from a workflow combining a text-to-speech model with independent music generation: here, Suno indicates that voice and music are produced as a single, coherent track. An Still-Imperfect Beta

Speech was tested for a month with a select group before the beta was opened to the entire Suno community.

The company still reports several unstable behaviors. Some accents can drift during a generation, while requested pauses can become excessively long or theatrical.

Suno is not currently providing any details about the model's architecture, its performance, the supported languages, or the level of independent control available over the voice and music after generation.

The beta is accessible from the latest version of the Suno application.