Voice and soundtrack in one take
Suno has launched Speech, a public beta that generates spoken voices from a script or a prompted description and can lay original background music under them. It runs on Suno’s web and mobile apps.
The company’s claim is about the architecture rather than the voice quality. “Today, we’re expanding what’s possible in Suno with Speech: the first audio model that generates voice and music together as one cohesive track,” said Jack Brody, Suno’s chief product officer, in the announcement. Conventional pipelines synthesise speech, then mix a bed under it. Suno says its model produces both in one pass.
The background music is optional and can be switched off with a toggle, leaving plain speech.
How it works, and what it will not do
There are two modes. Simple takes a description — the example Suno gives is “a pirate captain rallying his crew” — and writes and performs it. Advanced takes a script you supply, with settings for the voice’s gender, the speech style, and how much variation to allow between generations. A single piece of Speech output runs to a maximum of about eight minutes.
Brody is unusually direct about the state of it. “Beta really does mean beta,” he said. “Occasionally, British accents can wander off to Australia and back. Dramatic pauses may be very dramatic. You will almost certainly discover uses for this that never occurred to us.”
Why a music company is selling speech
Suno is a music generator, and its music generator has attracted a series of lawsuits. Expanding into spoken audio widens the product beyond the part of the catalogue that is being litigated, and into a market where the incumbents are general-purpose voice companies rather than record labels.
It also lands in a crowded fortnight. ElevenLabs released a speech architecture that accepts stage directions on 30 September. Microsoft shipped a streaming transcription model and two voice models on Wednesday, pricing text-to-speech at $22 per million characters. Fermion Research opened a transcription model small enough to run on a laptop.
Suno’s differentiator is the one thing none of those do: a voice and the music under it generated as a single object rather than assembled from two. Whether that matters depends on whether anyone wants the soundtrack. The toggle that turns it off is, in that sense, the most interesting control in the product.
