// AI NEWS · MODEL RELEASE

Google DeepMind Launches Lyria 3.5 in Flow Music, Upgrading AI Music Vocals, Lyrics, and Creative Control

The new music model rolled into Google Flow Music on July 29, 2026 as a free upgrade — richer melodies, cleaner vocals, better lyrics, and direct control over tempo and duration.

2026-07-31 · by Moe Ameen

What happened

Google DeepMind released Lyria 3.5, the newest version of its AI music generation model, on July 29, 2026, rolling it into Google Flow Music — the company's dedicated web and iOS music studio at flowmusic.app. It replaces the previous engine as the default, and for existing Flow Music users it arrived as a free upgrade with no new subscription tier and no waitlist.

Google describes the improvements in four areas. Musicality: richer, more complex melodic structures that sound more natural across a full track, addressing a common criticism of earlier releases that melodies felt mechanical after the opening section. Lyrics: higher-quality words with better prompt adherence and structural awareness. Vocals: more realistic, emotionally nuanced singing with improved pronunciation. Creative control: users can now steer tempo and duration more directly. As with all Lyria output, tracks are watermarked with SynthID so they can be identified as AI-generated.

The release is the third Lyria milestone of 2026. Lyria 3 landed in February, Lyria 3 Pro followed in late March with tracks up to roughly three minutes, and Lyria 3.5 now refines core output quality — a cadence that signals Google is treating AI music as a standing product line rather than a research demo. The same Lyria model line also powers music features in Gemini, YouTube Shorts' Dream Track, Google Vids, and Google AI Studio.

The timing sits against an unsettled backdrop for AI music. Rivals Suno and Udio remain in litigation with major labels over training data, and platforms are increasingly moving to badge or de-monetize AI-generated tracks. Google's licensed-and-watermarked positioning is part of how it distinguishes Lyria in that environment; creators should still plan disclosure for anything they monetize.

Why it matters for creators

  • Better vocals and lyrics lower the bar to a usable original track. A Lyria 3.5 song can now front a post, not just sit under B-roll — which makes AI music a viable hook, intro, or jingle for short-form.
  • Direct tempo and duration control means you can score to a specific slot. Matching a track to a 30-second Reel or a 15-second Shorts hook gets faster and less trial-and-error.
  • It is a free upgrade inside Flow Music. Existing users get the new model at no extra cost, so the practical barrier to trying AI music for content just dropped.
  • SynthID watermarking is the disclosure trade-off. Every track is flagged as AI-generated, which matters as platforms badge or throttle AI music — plan your labeling and keep a licensed option for monetized posts.
  • The launch itself is a content topic. "What Lyria 3.5 changes for creators" is exactly the timely, high-search subject that earns reach right now while competition for the keyword is thin.

How to act on this with Kompozy

There are two ways to act on this news today, and Kompozy — a full AI content generation and multi-platform publishing engine — covers both. The first is to make the launch your topic. Lyria 3.5 is a fresh, searched-for subject, and Kompozy turns one take on it into a week of coverage: a Blog Article explaining what changed, a brand-exact Carousel breaking down the four upgrades, native Text Posts per platform, Quote Graphics, and an Email Newsletter — all in your Persona Brief voice, then scheduled and published across nine destinations from one queue with Autopilot. Being early and useful on a thin keyword is how a small account gets cited and shared.

The second is to use the model. Generate a track in Flow Music, export it, and drop it into Kompozy as the audio bed for the formats that carry music — Naturalistic Video and Listicle Video over a portrait clip, Clipped Shorts for a longer piece, or a Persona Frames avatar composite — with captions burned in and per-platform reframing to 9:16, 1:1, and 16:9. Because every Lyria track is SynthID-watermarked and AI music sits in a shifting disclosure and monetization landscape, Kompozy's source-agnostic pipeline is the safety valve: if a monetized post needs a licensed track instead, you swap the soundtrack without rebuilding the video. React to the news and use the tool in the same pass.

Quick takeaways

  • Lyria 3.5 is live in Flow Music as a free upgrade — better vocals, lyrics, musicality, and tempo/duration control.
  • Every track is SynthID-watermarked; plan disclosure and keep a licensed option for monetized work.
  • Two plays: cover the launch as a timely topic, and use the tracks as audio beds for short-form video — Kompozy does both.

Frequently asked questions

When did Lyria 3.5 launch and where can I use it?

Google DeepMind launched Lyria 3.5 on July 29, 2026, inside Google Flow Music — the web studio at flowmusic.app and its iOS app. It became the default model as a free upgrade for existing users, and the broader Lyria line also powers music in Gemini, YouTube Shorts, and Google Vids.

What is new in Lyria 3.5 versus earlier versions?

Google cites more natural, complex melodies (musicality), higher-quality lyrics with better prompt adherence and structure, more realistic and emotionally nuanced vocals with cleaner pronunciation, and easier control over tempo and duration. It is a quality upgrade to the existing model, not a new product or tier.

Can I use Lyria 3.5 music in monetized videos?

You can, but every track carries a SynthID watermark identifying it as AI-generated, and platforms increasingly badge or de-monetize AI music. Plan disclosure, check each platform's policy, and for high-stakes commercial posts keep the option to swap in a licensed track — Kompozy's pipeline lets you change the soundtrack without rebuilding the video.

Related news

← All AI news · Get started →