Google introduced Lyria 3.5 on July 29, positioning it as an update to its music generation model with claimed gains in melody structure, lyric quality and vocal realism. The company rolled the model into Flow Music, its AI song creation tool at flowmusic.google, without a separate technical report or paper.
Google’s post lists four areas of improvement: richer melodic structures, better prompt adherence for lyrics, more emotionally expressive vocals with cleaner pronunciation, and finer control over tempo and song length. Each claim comes from Google’s own description alongside a demo clip. The company did not publish benchmark scores, listener evaluations, or comparisons against prior Lyria versions or rival music generators such as Suno or Udio.
AI-generated music remains legally contested industry-wide. Record labels have sued rival AI audio startups over the use of copyrighted recordings in training data, and questions about consent and compensation for musicians have followed every major music-model launch. Google’s post does not address what data trained Lyria 3.5, how artists are compensated, or how the model handles copyrighted material in its outputs. For labels and streaming platforms, Lyria 3.5 lands in a fast-moving category where marketing claims still outpace independent verification.
Google published the Lyria 3.5 update on its blog on July 29, 2026.
