AI music news
The few AI music updates worth knowing—what changed, why it matters, and the original sources.
Filter by topic, then keep scrolling. Every brief links to its sources.
4 updates
ElevenLabs released Music v2.5 as the new default model for prompted and reference generation in ElevenCreative. The company says the update produces richer melodies, more natural-sounding instruments and the biggest improvements in R&B, soul, hip hop, rock, metal, orchestral and cinematic music. Users can generate from written prompts or guide results with a reference track.
Google says Lyria 3.5 is rolling out in Flow Music with improved melodic complexity, lyric quality and prompt adherence, more expressive vocals and better pronunciation. The model also adds easier control over tempo and output duration. For creators, the release focuses on both generation quality and practical steering, but the claims currently come from Google's launch announcement rather than independent testing.
Mureka updated its API with a new Music Transcription endpoint, start and end controls for soundtrack generation, and an enhanced lyrics-generation model. The July 9 changelog groups the additions as one API and model update, but does not publish benchmark results or quantify the lyrics model’s improvement. For developers, the release adds a new analysis workflow and finer timing control when generating music for existing media.
Stability AI released Stable Audio 3.0 as a four-model family trained on licensed data. Three smaller variants arrived as open weights, while Large became available through the API and enterprise self-hosting; the family supports variable-length generation, audio editing and tracks beyond six minutes, with one Small model aimed at on-device composition. For creators and developers, this broadens deployment and editing options, though license terms differ by organization size.
