StepAudio 3 Music | EmpirioLabs AI Docs

StepFun · Audio Generation

POST /v1/audio/generations

Writes complete songs from a style description, with vocals, arrangement, and mixing, or instrumentals and covers.

At a glance

Field	Value
Model id	`stepaudio-3-music`
Model release date	2026-09-08
Input modalities	Text, Audio
Output modalities	Audio
Context window	-
Weight precision	-
Region	International
Features	music_generation, audio_generation, audio_in, multilingual
Native inference	No
New	Yes
Supported endpoints	`POST /v1/audio/generations`
Alternate model ids	`stepaudio-3-music-preview`, `stepfun/stepaudio-3-music`

Pricing

Charge	Spec	Rate
Music generation	per generated second	$0.0018

Example request

curl https://api.empiriolabs.ai/v1/audio/generations \
  -H 'Authorization: Bearer $EMPIRIOLABS_API_KEY' \
  -H 'Content-Type: application/json' \
  -d '{"model": "stepaudio-3-music", "prompt": "warm jazz piano", "duration": 8}'

Parameters

Parameter	Type	Required	Default	Description
`prompt`	string	yes	-	Style description: genre, vocals, mood, instruments, and key.
`lyrics`	string	no	-	Optional lyrics. Section tags such as [Verse], [Chorus], and [Outro] shape the structure. Left empty, the model writes them. Required for a cover or a vocal arrangement.
`task`	enum	no	`"text_to_music"`	Write a new song, cover a reference song, or build an arrangement around a dry vocal. · Allowed: `text_to_music`, `music_cover`, `vocal_to_music`
`instrumental`	boolean	no	false	Generate a track with no vocals. Cannot be combined with lyrics.
`song_audio`	string	no	-	Base64 reference song. Required for a cover.
`vocal_audio`	string	no	-	Base64 dry vocal with no accompaniment. Required for a vocal arrangement.
`response_format`	enum	no	`"mp3"`	Output audio format. · Allowed: `mp3`, `wav`, `flac`, `opus`, `pcm`
`temperature`	number	no	`0.85`	Sampling temperature. Higher is more varied. · Range: 0.01 – 2
`top_p`	number	no	`0.92`	Nucleus sampling probability mass. · Range: 0 – 1
`lyrics_rewrite`	boolean	no	false	Let the model rewrite the lyrics you supplied.

Notes

Describe the style and the model writes and performs a complete song. Supply lyrics or let the model write them, set instrumental for a track with no vocals, or attach a reference song for a cover or a dry vocal to build an arrangement around. Section tags such as [Verse], [Chorus], and [Outro] shape the structure. Output length is chosen by the model and is typically one to three minutes, so it is not set by a parameter and two runs of the same prompt can differ. Billing follows the generated length. Instrumental and lyrics cannot be combined. This model is in preview and its capabilities may change.

Machine-readable schema: GET https://api.empiriolabs.ai/v1/models/stepaudio-3-music.