![\"music](\"https:\/\/www.digitalmusicnews.com\/wp-content\/uploads\/2026\/05\/stability-ai-releases-stable-audio-3-trained-on-authorized-sources-1024x576.png\" "\\"music\\"")
<\/p>\n
Photo Credit: Stability AI<\/p>\n<\/div>\n
Music Stability AI launches Stable Audio 3.0, AI music models trained entirely on licensed data and able to generate tracks of over six minutes in length.<\/h2>\n
Stability AI has announced the launch<\/a> of Stable Audio 3.0, a family of four new AI music models that the company promises are trained entirely on licensed data. These new models can generate tracks of more than six minutes in length. Three of the four models are \u201copen-weight,\u201d meaning they are free to download and build upon.<\/p>\n The launch represents a major step up from the company\u2019s Stable Audio 2.0, which launched in 2024 with a maximum generation length of only three minutes.<\/p>\n \u201c3.0 Small generates up to two minutes, compared to 11 seconds from Stable Audio Open Small and 47 seconds from Stable Audio Open. 3.0 Medium and 3.0 Large generate more than six minutes,\u201d the company explains.<\/p>\n \u201cTo our knowledge, 3.0 Small is the only model capable of full music composition on-device. For the first time, on-device and offline audio generation isn\u2019t limited to short samples; it can produce complete musical tracks.\u201d<\/p>\n \u201cMusic has always evolved through the collective creativity of its community,\u201d the company added. \u201cRemix culture, interpolations, and mashups are how artists build on each other\u2019s work and push the art form forward. [\u2026] Generative audio will be no different. We want to foster the same kind of community-driven innovation in audio that we sparked in image generation with the launch of Stable Diffusion.\u201d<\/p>\n Under Stability AI\u2019s Community License, creators will also own their outputs and can distribute and commercialize them freely. The company added that, to its knowledge, \u201cother open music models either restrict commercial use or carry the risks associated with being trained on unlicensed music.\u201d<\/p>\n Organizations with more than $1 million in annual recurring revenue will require an enterprise license for commercial use, which Stability AI says also includes legal indemnification.<\/p>\n<\/section><\/div>\nA research paper published alongside the launch<\/a> says that the new models are trained on a combination of licensed audio from production library AudioSparx, comprising more than 800,000 audio files, and Creative Commons recordings from Freesound.<\/h4>\n