Google launches Lyria 3.5 for song generation in Flow Music

Google launched Lyria 3.5 on July 29, bringing its newest music-generation model into Google Flow Music. The release focuses on higher-quality lyrics, more expressive vocals, richer musical structure, and more direct control over tempo and duration.

Key takeaways

• Lyria 3.5 is available in Google Flow Music now.
• Google says the model improves musicality, lyrics, vocal realism, pronunciation, prompt adherence, tempo control, and duration control.
• Independent coverage reports tracks can run from 30 seconds to 3 minutes.
• The Decoder reports a new Selective Section Painting workflow for editing parts of a track without restarting.

What Lyria 3.5 adds

Google frames Lyria 3.5 as a quality and control update rather than a simple interface change. The official announcement says the model creates richer melodic structures, produces higher-quality lyrics, follows prompts better, and improves vocal expression and pronunciation.

That matters because AI music tools often fail at the same points: generic lyrics, unstable vocals, muddy pronunciation, weak structure, and little control after generation. Lyria 3.5 appears aimed at those failure modes.

Where it is available

The launch surface is Google Flow Music. Google says users can try Lyria 3.5 there now, and independent hands-on coverage from Gigazine reports that generation was available through the Flow Music web app.

Gigazine also reports that users can generate songs from text prompts and image inputs, then export audio in common formats including MP3, M4A, and WAV. LinkLoot has not independently tested those export paths, so treat them as third-party hands-on context rather than a platform guarantee.

Why creators may care

For creators, the important change is the combination of model quality and editability. The Decoder reports that Selective Section Painting lets users revise specific parts of a track or expand short melodies into full songs without discarding the whole result.

That is a workflow change. Instead of prompt, wait, reject, and regenerate, a creator can work closer to arrangement and revision. If the controls hold up in real projects, Lyria 3.5 becomes more useful for drafts, social cuts, demo vocals, and soundtrack exploration.

Limits to watch

Google's announcement does not provide a detailed training-data disclosure for Lyria 3.5. The Decoder notes that Google previously said Lyria 3 used materials it had rights to use, but Google had not provided more detail about Lyria 3.5 training data at publication time.

Rights, commercial-use terms, regional availability, credit limits, and export behavior should be checked inside Flow Music before creators build a paid workflow around it.

Source check

• Google announcement confirms the Lyria 3.5 launch, Flow Music rollout, and claimed improvements to musicality, lyrics, vocals, tempo, and duration controls.
• The Decoder corroborates the launch and reports the 30-second to 3-minute track range plus Selective Section Painting.
• Gigazine provides independent hands-on context for Flow Music access, image-to-song prompting, and export options.