Suno adds Speech beta: spoken text and matching background music generated as a single audio track

Suno, best known as an AI music generator, has added a feature called Speech that produces spoken text and matching background music together as a single audio track, The Decoder reported on October 2, 2026. The feature is in beta.

How Speech works

Users type in an idea or supply written text, then describe the voice and music style they want. The model generates both the voice and the accompanying sound, and the result is delivered as one audio track rather than separate voice and music files.

According to Suno, the feature suits poems, meditations and bedtime stories.

Testing and known issues

Suno product chief Jack Brody says the company tested Speech with a small group for a month before launch. Suno acknowledges the beta still has bugs. One example it gives: a requested British accent can sometimes sound Australian.

What Suno has not disclosed

The source report does not include:

• The name or version of the model behind Speech, or whether it is a new model or an extension of Suno's existing music model
• Parameter count, architecture or training cutoff
• Pricing or plan availability (pricing not yet disclosed)
• Limits on audio length, supported languages or available voice options
• Benchmark results

Suno also has not said how the model was trained.

Legal context

The training-data question matters because AI music generators face criticism over potential copyright infringement. Major record labels have already sued Suno. According to The Decoder, a Munich court recently ruled against the startup and rejected fair use as a justification for using copyrighted data. The report does not give details of that ruling beyond this.

What this means

Speech moves Suno beyond music into spoken-word content, and the notable design choice is producing voice and score as one track. Creators making meditations, narrated poems or bedtime stories would otherwise need a text-to-speech tool, a music source and an editor to mix them. Suno's approach removes that step, though it also removes the ability to adjust the voice and music levels independently. That trade-off has not been addressed in the available information.

The launch puts Suno in a space already occupied by dedicated voice-synthesis providers. Whether Speech competes on voice quality is impossible to judge without a model name, samples or evaluations. The reported accent errors suggest voice control is still imprecise.

The unresolved copyright situation is the larger risk. Suno has not disclosed training details for Speech while it faces label lawsuits and an adverse Munich ruling on fair use. Any new model built on similar data practices carries the same legal exposure, and commercial users should watch how those cases develop before relying on the output.