Google DeepMind Launches Lyria 3.5 in Flow Music | AI Music

Google DeepMind released Lyria 3.5 on July 29, 2026, rolling it out immediately inside Google Flow Music, the company’s web-based AI music creation studio. The update targets four areas that have long separated Google’s offering from market leaders Suno and Udio: musicality, lyrics, vocals and creative control over tempo and duration.

It’s the third significant Lyria release in roughly five months — Lyria 3 arrived in February 2026 with 30-second clip generation, Lyria 3 Pro followed in late March with tracks up to three minutes, and now Lyria 3.5 refines the core output quality across the board. Google did not publish numerical benchmarks alongside the announcement.

What Actually Changed

On the musicality side, the model is designed to produce more complex melodic structures that feel less mechanical. The lyrics engine received what Google describes as improved prompt adherence and structural awareness — meaning the model is more likely to respect the shape and intent of what you describe, rather than generating plausible-sounding but structurally incoherent verses. Vocals, historically a weak point for Lyria compared to Udio, now aim for greater emotional nuance and better pronunciation. Users also gain more direct control over tempo and the duration of generated tracks, two parameters that previously required workarounds.

Google did not confirm in the announcement whether Lyria 3.5 replaces Lyria 3 Pro on the Gemini app or developer API, leaving some ambiguity about which surface gets which model version going forward.

Where This Sits in the AI Music Wars

Suno currently leads the consumer AI music market by a wide margin — roughly 2 million paid subscribers and $300 million in annual recurring revenue as of mid-2026 — and its v5.5 release added voice cloning and a built-in digital audio workstation called Suno Studio. Udio has carved out a different niche, attracting producers who want finer creative control, remixing tools and stem downloads. Both platforms charge $10 per month for their core paid plans.

But both competitors carry legal baggage. Suno settled with Warner Music in November 2025, and active litigation from Sony Music and Universal Music Group was still ongoing as of July 2026. Google, by contrast, trained Lyria exclusively on licensed material. Every track generated through Flow Music also carries a SynthID watermark — an embedded provenance signal that survives common audio transformations and is increasingly being adopted as an industry standard as streaming platforms build out AI-detection systems.

Google’s structural advantages — legal cleanliness, SynthID and deep integration with the Gemini ecosystem — were already in place before this update. What Lyria 3.5 adds is a direct push on audio quality, particularly in the vocal and lyric departments where Udio has held an edge. Whether the improvements close that gap meaningfully will depend on how the model performs on real-world prompts, not marketing copy.

Why Students Should Pay Attention

Pricing is where Google’s pitch to students gets genuinely compelling. Flow Music’s core tier is free with a standard Google account — no credit card required. The AI Pro plan, which unlocks 1,000 credits per month, is priced at $19.99 per month for general users, but college students can access it free for one year through Google’s education program. That puts the full Lyria 3.5 experience within reach for students who would otherwise max out on free-tier credit limits quickly.

The improvements to prompt adherence and structural awareness have practical implications beyond casual use. A film or media production student who needs a score bed for a short film can now describe a scene’s emotional arc and receive a track with intentional structure, rather than something that sounds like disconnected audio. Marketing and communications students can generate podcast beds or branded jingles that would previously have required hiring a composer or licensing stock music.

The legal distinction matters too, especially for anyone with commercial intent. YouTube channels, Etsy shops, freelance reels and student-run brands all carry lower legal risk when the underlying AI tool was trained on licensed data — which, among the major AI music generators available today, currently describes Lyria and not its two largest competitors.

The Takeaway

Lyria 3.5 doesn’t dethrone Suno’s market dominance or match Udio’s producer-grade toolkit overnight. What it does is narrow the quality gap in vocals and lyrics while keeping Google’s core advantages intact: licensed training data, SynthID watermarking, and a free tier that costs nothing to try. For students creating anything with a commercial dimension — or anyone who wants to avoid ending up as a footnote in a music label’s next copyright lawsuit — Flow Music is worth a serious look right now.

Source: Google DeepMind