Stable Audio 3.0

Audio modelCard summarized

Our most advanced audio generation model, capable of generating up to 6-minute, 44.1 kHz stereo compositions.

Source: Stable Audio 3 paper (arXiv 2605.17991, Stability AI), Stable Audio 3.0 launch post and API reference

Use it via API

$0.08 / song

Code example Compare

Where to call it

Stability AI Platform APIstable-audio-3

Supports

No feature list in the maker's docs.

Full feature support

Good for

Full instrumental music tracks up to about six minutes
Individual instruments and stems (TrackType: Instrument)
Samples and sound effects (TrackType: SFX)

Overview

Stability AI · 02Large

Stable Audio 3

Our most advanced audio generation model, capable of generating up to 6-minute, 44.1 kHz stereo compositions.

Typical price

$0.08 / song

Any generation, 1 s to 6 min 20 s

Max length: 380 s
Languages: English prompts only
Modes: text-to-audio (music and sound effects), audio-to-audio
Released: 20 May 2026

Stable Audio 3.0 is Stability AI's family of latent diffusion models for music and sound effects, released on 20 May 2026. The Large model is the one offered on Stability's API and to enterprises; Small and Medium are open weights. It generates variable-length 44.1 kHz stereo audio up to 6 min 20 s from text, transforms uploaded audio and inpaints or continues existing recordings. Stability says it was trained on fully licensed data, and the API charges a flat 8 credits ($0.08) per generation. paper; launch post; API reference

stable-audio-3

Typical price: $0.08 / song
Max length: 380 s (6 min 20 s)
Languages: English prompts only
Weights: Closed
Released: 20 May 2026

text-to-audio (music and sound effects)
audio-to-audio
audio inpainting
continuation
stems via prompt tags