Google Lyria 3.5 vs Suno AI: What Changed in AI Music Generation

Google DeepMind released Lyria 3.5 in Flow Music, its third major AI music update since February. Here is what actually changed, and how it compares to Suno AI on paper.

*Some links in this article are affiliate links. If you sign up through one, ShabelleHub may earn a commission at no additional cost to you. See our Affiliate Disclosure for more information.*

Google DeepMind released Lyria 3.5 on July 29, 2026, inside Google Flow Music. It's the third significant Lyria milestone since February \u2014 after Lyria 3 in February and Lyria 3 Pro in March \u2014 which signals Google is treating AI music generation as an active product line rather than a side experiment.

Editorial note: ShabelleHub has not run Lyria 3.5 and Suno AI side by side. The comparison below is based on each company's own stated features and independent reporting, not hands-on testing. We're publishing this as a news-and-context piece rather than adding Lyria to our directory with a star rating, since we only assign ratings after direct testing.

What Actually Changed in Lyria 3.5

According to Google's announcement, the update focuses on four areas:

• Musicality \u2014 richer melodic structures and chord progressions described as less repetitive and more intentional-sounding than earlier Lyria versions.
• Vocals \u2014 improved emotional nuance and pronunciation, aimed at making lead and backing vocals sound less synthetic.
• Lyrics \u2014 tighter adherence to prompts and better handling of song structure (verses, choruses, bridges staying coherent across a full track).
• Creative control \u2014 direct control over tempo, plus separate controls for vocals, drums, and bass, alongside track-length options. Tracks can now run from 30 seconds up to three minutes.

The update is live now for all Google Flow Music users at no additional cost.

Where Lyria Lives: Google Flow Music

Lyria 3.5 isn't a standalone product \u2014 it's the model powering Google Flow Music, the platform where users actually generate songs. Flow Music has an unusual lineage: it began as the open-source Riffusion project in 2022, relaunched as ProducerAI in 2025, and became an official Google product in February 2026 when Lyria 3 shipped inside it. Flow Music offers a free tier alongside a paid subscription reported to start around $6/month, though pricing details are worth confirming directly on Google's site before signing up, since subscription tiers change.

How It Compares to Suno AI, on Paper

Suno AI has a head start as one of the most established text-to-song generators, and it remains our top-rated pick in the Audio category. Both tools generate full songs \u2014 vocals, lyrics, and instrumentation \u2014 from a text prompt, and both are aimed at the same use case: fast, complete song generation without needing to play an instrument.

Where they differ, based on what's publicly documented:

• Track length: Suno's outputs are typically shorter-form; Lyria 3.5 explicitly supports up to three minutes.
• Creative control: Lyria 3.5's separated controls for tempo, vocals, drums, and bass are more granular than what Suno exposes in its standard interface.
• Distribution: Suno is a standalone product built specifically for song generation. Lyria 3.5 lives inside Flow Music, a broader Google-owned platform, and is also available to developers through Vertex AI and the Gemini API.
• Commercial rights: Suno requires a paid plan for commercial usage rights, a limitation we've noted in our Suno review. Google has not published equally specific commercial-use terms for Flow Music output at the time of writing \u2014 worth checking directly before using generated tracks commercially.

The Copyright Question Nobody's Ignoring

Any AI music tool trained on existing recordings sits in contested legal territory right now, and Lyria is no exception. Google is currently defending itself in a lawsuit brought by independent artists over how the Lyria model family was trained. Reporting on the case indicates Google has argued that uploading music to YouTube constitutes a license broad enough to cover AI training use \u2014 an argument that, if a court accepts it, would mean the YouTube terms of service users agreed to years ago effectively became a training license they didn't realize they'd signed.

The case is unresolved. We're not taking a position on how it should be decided \u2014 just flagging that it exists, since it's directly relevant to anyone deciding whether to use Lyria-generated music commercially. Google has stated Lyria doesn't mimic specific artists and that outputs are marked with SynthID to indicate AI generation.

Where This Leaves Things

Lyria 3.5 is a real, meaningful update \u2014 longer tracks, more granular control, and vocal quality that multiple outlets independently describe as a step up from earlier versions. Whether it's *better* than Suno for a given project depends on what you're optimizing for: Suno's simplicity and established track record, or Lyria's longer clips and more granular controls, bundled inside Google's broader creative ecosystem.

We'll consider adding Lyria to the ShabelleHub directory with a full rating once we've had a chance to test it hands-on, consistent with our review methodology.

*This article is part of ShabelleHub's ongoing coverage of AI tool releases.*

Frequently Asked Questions

📬 Get the latest AI tool reviews

Expert picks and comparisons, weekly. No spam.