{"rewrite":{"id":"r_604703d00e49465d88c8ceba","clusterId":"c_7fd0f47a22314b82b2158390","slug":"stability-ai-releases-stable-audio-3-0-with-6-minute-music-generation","model":"deepseek-v4-flash","headline":"Stability AI Releases Stable Audio 3.0 With 6-Minute Music Generation","summary":"Stability AI released Stable Audio 3.0, a new series of music generation AI models trained on licensed data from Universal Music Group and Warner Music Group. The models are open-weight and allow commercial use. The lineup includes models capable of generating songs over 6 minutes, with the Small variant supporting full on-device composition for the first time.","whyItMatters":"Stable Audio 3.0 is the first open-weight music generation model trained entirely on licensed data that permits unrestricted commercial use, addressing a central legal and ethical dispute in the AI music space.","webCardHtml":"\u003cp\u003eStability AI has released Stable Audio 3.0, a new generation of music and audio generation AI models. The company trained the models on fully licensed data through partnerships with Universal Music Group and Warner Music Group, and released them as open-weight models. Users own the outputs and can use them commercially without restriction. The model lineup includes three open-weight variants: Small SFX for sound effects, Small for songs up to 2 minutes, and Medium for songs over 6 minutes. A Large model is available via API. Stability AI says the Small model is the first capable of full song composition on device and offline. The release also introduces LoRA fine-tuning for audio generation models for the first time, along with audio inpainting features for editing specific segments or extending audio. For organizations with annual revenue over $1 million, an enterprise license provides legal indemnification. A new product suite for musicians is under development.\u003c/p\u003e\u003cul\u003e\u003cli\u003e\u003cstrong\u003e3.0 Small SFX\u003c/strong\u003e: Open-weight model specialized for sound effect generation\u003c/li\u003e\u003cli\u003e\u003cstrong\u003e3.0 Small\u003c/strong\u003e: Open-weight model capable of generating songs up to 2 minutes; first model to support full song composition on device\u003c/li\u003e\u003cli\u003e\u003cstrong\u003e3.0 Medium\u003c/strong\u003e: Open-weight model capable of generating songs over 6 minutes\u003c/li\u003e\u003cli\u003e\u003cstrong\u003e3.0 Large\u003c/strong\u003e: API model capable of generating songs over 6 minutes\u003c/li\u003e\u003c/ul\u003e","blueskyPost":"Stability AI released Stable Audio 3.0, trained on licensed data from Universal and Warner. Open-weight models generate up to 6 minutes of music, with full on-device composition and commercial use allowed.","twitterPost":"Stability AI released Stable Audio 3.0, trained on licensed data from Universal and Warner. Open-weight models generate up to 6 minutes of music, with full on-device composition and commercial use allowed.","threadsPost":null,"newsletterBlurb":"Stability AI released Stable Audio 3.0, a new series of music generation AI models trained on licensed data from Universal Music Group and Warner Music Group. The open-weight models allow commercial use and include variants capable of generating songs over 6 minutes. The Small model is the first to support full song composition on device and offline.","attributionJson":"[{\"source\":\"GameBusiness.jp\",\"url\":\"https://www.gamebusiness.jp/article/2026/05/21/27005.html\",\"title\":\"Stability AI Announces 'Stable Audio 3.0' Capable of Generating Up to 6 Minutes of Music, Available for Commercial Use\"}]","lintFlagsJson":"[]","lintHits":0,"costUsd":0,"inputTokens":3742,"outputTokens":741,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1779826585,"createdAt":"2026-05-21T13:23:22.000Z","publishedAt":"2026-05-21T13:33:25.000Z","updatedAt":"2026-05-21T13:33:25.000Z"},"cluster":{"id":"c_7fd0f47a22314b82b2158390","canonicalTitle":"Stability AI、最大6分の音楽が生成できる「Stable Audio 3.0」発表。商用利用も可能","representativeArticleId":"a_3db4bda7190e8d3fcffa5c93","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"studios\":[],\"people\":[],\"type\":\"announcement\",\"domain\":\"music\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-05-21T11:15:04.000Z","lastSeenAt":"2026-05-21T11:15:04.000Z","updatedAt":"2026-05-21T13:33:25.000Z"},"attribution":[{"source":"GameBusiness.jp","url":"https://www.gamebusiness.jp/article/2026/05/21/27005.html","title":"Stability AI、最大6分の音楽が生成できる「Stable Audio 3.0」発表。商用利用も可能"}],"entities":{"anime_titles":[],"manga_titles":[],"studios":[],"people":[],"type":"announcement","domain":"music","is_roundup":false},"keyFacts":["Stability AI released Stable Audio 3.0, a series of music generation AI models trained on licensed data from Universal Music Group and Warner Music Group.","The models are open-weight and allow unrestricted commercial use, with users owning the outputs.","The lineup includes three open-weight variants: Small SFX for sound effects, Small for songs up to 2 minutes, and Medium for songs over 6 minutes, plus a Large model via API.","The Small model is the first capable of full song composition on device and offline.","The release introduces LoRA fine-tuning for audio generation models and audio inpainting features for editing specific segments or extending audio."]}