DeepMind Ships Lyria 3.5 Music Model Inside Google Flow
*Google’s newest audio generator claims gains in musicality, lyrics, vocals, and user control over track creation.*
Google DeepMind and Google both published announcements on 29 July 2026 stating that Lyria 3.5 is now live in Google Flow Music. The model is presented as the successor to earlier Lyria versions and is positioned for users who want to generate full tracks rather than short clips.
The release notes list four areas of reported progress: musicality, lyrics, vocals, and creative control. No specific metrics, training details, or benchmark scores appear in either post. The text simply states that the changes let users “craft richer tracks.”
Both announcements use nearly identical wording and carry the same date stamp. No independent verification or third-party testing is referenced.
Limited information released
The source posts contain no examples of generated audio, no comparison samples against prior versions, and no statements on model size, latency, or licensing terms. Readers are left with the four bullet-point claims and a link to the product inside Flow Music.
Why it matters
Engineers and producers who already work inside Google’s music tools now have an updated generator to test. Whether the improvements are noticeable will depend on hands-on use rather than the announcement text. Until concrete samples or independent evaluations surface, the update remains an internal product claim rather than a demonstrated technical advance.
---
Sources:
No comments yet