Google Puts an AI Music Studio Inside Gemini with Lyria 3

Google dropped Lyria 3 into the Gemini app today. It’s DeepMind’s latest music generation model, and it lets you create 30-second songs with vocals, lyrics, and cover art from a text prompt. Or from a photo. Or from a video. You describe what you want, and Gemini builds a track.

The feature is rolling out globally to all Gemini users 18 and older, across eight languages. No paywall for basic access, though Google AI Plus, Pro, and Ultra subscribers get higher generation limits. It’s in beta.

This is the first time a major general-purpose AI assistant has shipped built-in music generation to a user base this size. Gemini has over 750 million monthly active users. Suno, the current market leader in AI music, has 12 million.

How It Works

You type a prompt like “a comical R&B slow jam about a sock finding its match” and Gemini generates a 30-second track with lyrics and cover art. The cover art uses Google’s Nano Banana image model.

Compared to earlier Lyria versions, the big change is that you no longer need to write your own lyrics. Lyria 3 generates them from the prompt. You can also adjust style, vocals, and tempo, or upload a photo or video and have Gemini compose something that matches the mood.

Google is also rolling out Lyria 3 to YouTube creators through the Dream Track feature, which means the same model will power both consumer and creator-facing music tools.

Every generated track carries a SynthID watermark. You can upload any audio file to Gemini and it’ll tell you whether it was made with Google’s AI. This is the same watermarking system Google uses for AI-generated images and video.

The Copyright Situation

This is where it gets complicated. Google says it trained Lyria 3 on content “YouTube and Google have the rights to use” and has been “very mindful of copyright and partner agreements.” But a 2024 Billboard report revealed Google initially trained earlier models on copyrighted major-label recordings before securing permissions.

If you type a specific artist’s name, Gemini treats it as “broad creative inspiration” rather than attempting direct imitation. Filters check outputs against existing recordings. But Google acknowledges the filters “might not get everything right” and asks users to report infringements. That’s not exactly airtight.

For context, both Suno and Udio spent most of 2024 and 2025 fighting major label lawsuits. Both have now settled. Warner struck a licensed-model deal with Suno, and Universal settled with Udio and moved it toward a licensed platform. Google is trying to skip that lawsuit phase entirely by claiming it had the rights from the start. Whether that holds up remains to be seen.

How It Compares to Suno and Udio

Lyria 3 generates 30-second clips. Suno generates full songs over 90 seconds in under a minute. Udio produces similar-length tracks with what users describe as cleaner vocals and better instrumental separation.

The quality gap matters. Suno’s v5 model delivers what reviewers call “streaming-ready audio” with solid vocal realism and consistent mixes. Udio, built by former Google DeepMind engineers (ironic given this announcement), is the preferred tool among professional producers for its granular controls and stem downloads.

Lyria 3 is simpler by design. No stem exports. No granular mixing controls. No song extensions beyond 30 seconds. The trade-off is accessibility: it lives inside an app that hundreds of millions of people already use, and it takes zero onboarding to try.

Here’s the market as it stands:

Platform	Users	Track Length	Pricing	Strengths
Suno	12M	90+ sec	Free tier / $10-$30/mo	Full songs, v5 quality, 67% market share
Udio	4.8M	90+ sec	Free tier / $10-$30/mo	Pro controls, stem downloads, vocal clarity
Lyria 3 (Gemini)	750M+ potential	30 sec	Free (beta)	Zero friction, photo/video input, massive reach

Early Reactions

The response has been split.

Android Central called it the point “where everyday AI gets interesting,” arguing that embedding music generation into a mainstream assistant brings generative audio to a much wider audience than dedicated tools ever could.

Tom’s Guide tested it by turning a to-do list into a punk rock anthem. The novelty landed, though 30 seconds is short enough to feel like a demo rather than a product.

The Register was less kind, calling the outputs “soulless, 30-second musical AI slop” and mocking Google’s claim that generating music from a text prompt counts as “self-expression.” Their take: if you contributed nothing but a sentence, calling the result yours is a stretch.

The music industry response has been cautious. Google highlighted Warner Music CEO Robert Kyncl’s past praise of YouTube’s “partnerships approach,” but Music Ally noted he spent 12 years at YouTube before joining Warner, so his endorsement comes with context.

What This Actually Means

Lyria 3 is not going to replace Suno or Udio for anyone who cares about music production. It’s too short, too simple, and too limited in its output controls. If you’re making songs to publish or share, the dedicated tools are better at every level.

What Lyria 3 does is normalize AI music generation for a general audience. Most people have never opened Suno. Most people have never heard of Udio. But hundreds of millions of people open Gemini every day. If even a small percentage of them try music generation and get hooked on the concept, the downstream demand for better tools grows.

Google’s playbook here is the same one they’ve run with every Gemini feature: ship it free to the widest possible audience, get people used to the interaction pattern, then improve quality and upsell later. Lyria 3 is a top-of-funnel play, not a finished product.

The 30-second limit is both a technical constraint and a smart business decision. It’s enough to be fun without threatening the music industry enough to trigger lawsuits. Google clearly watched what happened to Suno and Udio and decided to start conservative.

Whether the quality improves fast enough to matter is the real question. DeepMind has the talent and the compute. But Suno and Udio have two-year head starts, dedicated user bases, and fresh label deals that give them legal cover Google is still working to establish.

For now, Lyria 3 is a toy inside a very large distribution channel. That combination has a way of becoming something more.

---

Sources:

• TechCrunch: Google adds music generation to Gemini
• 9to5Google: Gemini app rolling out music generation with Lyria 3
• Google Blog: Create custom soundtracks with Lyria 3
• Music Ally: Google adds Lyria 3 to Gemini
• Music Business Worldwide: Google launches Lyria 3
• The Register: Gemini will now generate musical slop
• Android Central: Where everyday AI gets interesting
• Tom’s Guide: Lyria 3 hands-on
• Suno vs Udio market data