Google has released its latest AI music model, Lyria 3.5, now available in Google Flow Music. The new model zeroes in on melody, lyrics, and vocal quality — and it shows. Generated melodies are richer, lyrics stick closer to your prompts, and the vocals sound more accurate and emotionally natural.
Users can specify genre, instruments, vocals, tempo, and song length. Lyria can generate up to 3-minute songs. Flow Music also allows partial lyric rewriting, translation, beat changes, and extending specific sections without regenerating the entire track.
The previous Lyria 3 Pro already supported full 3-minute songs. Lyria 3.5 doesn't push length further; instead, it improves output quality and control precision. Every generated song gets a SynthID invisible watermark to identify AI-created or modified music.