Google launches Lyria 3.5 AI music model in Flow Music

Google Lyria 3.5 AI music generation Flow Music SynthID watermark

Google announced the release of Lyria 3.5, its next‑generation AI music‑generation model, on 29 July 2026. The model is deployed directly within the Google Flow Music service, where users can create full‑length songs up to roughly three minutes. Lyria 3.5 expands on earlier versions by delivering richer melodic structures, higher‑quality lyrics and more realistic, expressive vocals. It also gives creators granular control over tempo and track length and accepts both text and image prompts.

Technical details note that the model was trained on a large audio corpus using Google’s own TPU hardware together with the JAX and ML Pathways frameworks. All generated audio is embedded with a SynthID watermark so that AI‑produced tracks can be identified later. The rollout is global, with the model available to anyone using Flow Music.

In parallel, major record‑label groups—including Sony Music, Universal Music and Warner Music—have formed a coalition seeking to bar AI‑generated songs from worldwide chart listings, citing concerns over authorship and market impact.

Entities: DeepMind · Flow Music · Google · Google Flow Music · Google LLC · Lyria 3.5 · Sony Music · SynthID

Claims

What the coverage asserts, and how well corroborated each claim is across sources.

[● 2 SOURCES] The model includes a SynthID watermark to identify AI‑generated audio. (Google)
[● 3 SOURCES] Lyria 3.5 improves musicality, lyric quality, and vocal realism compared with earlier versions. (Google)
[○ 1 SOURCE] The Flow Music Spaces upgrade is intended to position Google against AI‑music competitors Suno and Udio. (Google)
[○ 1 SOURCE] Google updated Flow Music Spaces with AI tools for song editing, stem splitting, image and video generation, and lyric creation within a single workspace. (Google)
[● 4 SOURCES] Lyria 3.5 is integrated into the Google Flow Music platform and available to users. (Google)
[● 3 SOURCES] Google announced Lyria 3.5 on July 29, 2026. (Google)
[● 3 SOURCES] Google DeepMind released the Lyria 3.5 music‑generation model. (Google DeepMind)
[○ 1 SOURCE] Lyria 3.5 can generate songs up to about three minutes long and supports text and image prompts. (Google DeepMind)
[● 3 SOURCES] Lyria 3.5 improves musicality, lyric quality and vocal realism over previous Lyria models. (Google)
[● 4 SOURCES] Google released the Lyria 3.5 music‑generation model. (Google)
[● 2 SOURCES] Users can control tempo and length of generated songs with Lyria 3.5. (Google)
[● 2 SOURCES] Lyria 3.5 outputs are watermarked with SynthID to identify AI‑generated audio. (Google)