Stability AI releases open-weights Stable Audio 3.0
stability ai released stable audio 3.0, generating six-minute tracks from licensed data with three open-weight models available for commercial use under a new community license.
stability ai launched stable audio 3.0, introducing a semantic-acoustic autoencoder architecture that allows for variable-length generation up to six minutes and 20 seconds, a significant increase over previous versions. the release includes four model variants: small sfx and small for short clips, medium for longer tracks, and a closed 2.7 billion parameter large model reserved for api and enterprise use. the three smaller models are released under the stability ai community license, permitting free commercial use for organizations earning less than one million dollars annually, and include support for inpainting and causal continuation. by training entirely on licensed data and securing partnerships with universal music group and warner music group, stability ai aims to mitigate copyright risks that currently plague competing open-source music models.