Just a month ago, Lyria 3 could only generate 30-second clips. Then on March 25, Google dropped Lyria 3 Pro and changed the game entirely. Three-minute full tracks, structure-aware composition that understands intros-verses-choruses-bridges, and a developer API to boot.
In our Suno vs Udio comparison, I wrote that "if Google enters the AI music market seriously, the landscape could shift." Well, it happened.
TL;DR
- Lyria 3 Pro: Google DeepMind's latest AI music model, generates up to 3-minute full tracks
- Structure-aware composition: specify intros, verses, choruses, and bridges via prompts
- 48kHz stereo high-quality audio output with automatic lyrics and vocals
- Image-to-music: feed a photo and get music matching its mood
- SynthID watermark: all outputs automatically tagged as AI-generated
- API: available in public preview via Gemini API, Vertex AI, and Google AI Studio
- Pricing: approximately $0.08 per song via API; Gemini subscriptions $19.99–$99.99/mo
- No artist mimicry: artist names in prompts are treated as "broad inspiration" only
From 30 Seconds to 3 Minutes: What Changed
Lyria 3, launched in February 2026, was limited to 30-second clips. Good quality, but too short for real-world use. Lyria 3 Pro breaks through that ceiling.
Key Upgrades
| Feature | Lyria 3 (Feb) | Lyria 3 Pro (Mar) |
|---|---|---|
| Track length | Max 30s | Max 3 min |
| Structure awareness | None | Intro/verse/chorus/bridge |
| Audio quality | 48kHz stereo | 48kHz stereo |
| Lyrics | Basic | Time-aligned (timing control) |
| Tempo control | None | Precise BPM setting |
| Image input | None | Image-to-music |
| API | Preview | Gemini API + Vertex AI |
Structure-aware composition is the biggest differentiator. You can specify things like "start with a 30-second synth intro, bring in the chorus at 1:00, add a bridge at 2:00" — and the model actually follows it.
Prompt Examples in Practice
Lyria 3 Pro works with text prompts. Here's what they look like:
Basic Prompt
"Acoustic guitar and piano-driven indie pop.
BPM 95. Female vocals. Theme: a rainy day at a café"
Structure-Specified Prompt
"Electronic pop, BPM 128.
0:00-0:15 synth pad intro
0:15-0:45 verse 1 - male vocals
0:45-1:15 chorus - energy up, add drums
1:15-1:45 verse 2
1:45-2:15 chorus repeat
2:15-2:45 bridge - piano solo
2:45-3:00 outro"
Image-to-Music
Feed a photo, get music that matches its mood. Sunset photo? Warm, nostalgic track. City skyline at night? Electronic vibes. It's a genuinely novel input modality for music generation.
Developer API: Gemini API + Vertex AI
This is the most important part for developers. Lyria 3 Pro is accessible via Gemini API and Vertex AI in public preview.
Access Methods
| Platform | Access | Best For |
|---|---|---|
| Google AI Studio | Web UI, immediate testing | Prototyping |
| Gemini API | REST API calls | App integration |
| Vertex AI | Enterprise environment | Production deployment |
| Gemini App | Mobile/web app | End users |
| Google Vids | Video editing tool | Auto BGM |
Pricing Structure
| Access Method | Price | Per Song | Daily Limit |
|---|---|---|---|
| API (Vertex AI) | Undisclosed (preview) | ~$0.08 | - |
| Gemini AI Plus | $19.99/mo | ~$0.067 | 10 songs/day |
| Gemini Pro | $29.99/mo | ~$0.05 | 20 songs/day |
| Gemini Ultra | $99.99/mo | ~$0.067 | 50 songs/day |
At ~$0.08 per song via API, Lyria 3 Pro is about 5x more expensive than Suno's subscription rate ($0.015/song). But the ability to automate via API and integrate with Google's ecosystem (Vids, YouTube) makes a direct price comparison misleading.
Suno vs Udio vs Lyria 3 Pro: The New Three-Way Race
The Suno vs Udio duopoly now has a third contender.
| Feature | Suno v4 | Udio | Lyria 3 Pro |
|---|---|---|---|
| Track length | 4 min | 15 min | 3 min |
| Structure control | Basic | Detailed | Detailed (time-aligned) |
| Vocal quality | Good | Excellent | Good |
| API | None (unofficial) | None | Official API |
| Image input | None | None | Supported |
| Watermark | None | None | SynthID |
| Per-song cost | ~$0.015 | ~$0.02 | ~$0.08 |
| Commercial use | Paid plans | Paid plans | Paid plans |
Lyria 3 Pro's biggest advantage is the official API and Google ecosystem integration. Suno and Udio have no official APIs, making automation difficult. On the other hand, track length and per-song cost favor the competition.
SynthID: An ID Card for AI Music
Every Lyria 3 Pro output is embedded with a SynthID watermark — inaudible to human ears but machine-detectable.
As we covered in our Google Search Live analysis, SynthID is Google's company-wide technology for identifying AI-generated content. Voice, images, and now music.
This matters for regulatory compliance. The EU AI Act mandates AI content labeling, and in music copyright disputes, determining whether a track is AI-generated is becoming a central issue.
Artist protection is also notable. If you name a specific artist in your prompt, Lyria 3 Pro won't mimic them — it treats the name as "broad inspiration" only.
Honest Assessment
What's Good
- 3-minute full tracks + structure awareness: Music that's actually usable
- Official API: The only option for integrating music generation into apps
- Google ecosystem: Natural integration with Vids, YouTube, Gemini
- SynthID + artist protection: Minimized copyright risk
What's Lacking
- 3-minute cap: Shorter than Suno (4 min) and Udio (15 min)
- Price: ~$0.08/song is 5x more expensive than competitors
- Still in preview: API pricing not finalized, stability unverified
- Mimicry restrictions: Difficult to precisely reproduce specific styles (by design)
What Developers Can Try Right Now
- Test Lyria 3 Pro in Google AI Studio: ai.google.dev
- Try Gemini API code integration
- Experience automatic BGM generation in Google Vids
- Run comparison tests against existing Suno/Udio workflows
The AI music market is shifting from a Suno-Udio duopoly to a three-way race including Google. The API accessibility is Lyria 3 Pro's unique advantage. If you've been wanting to add music generation to your app, now's the time to start.
Have you tried making music with AI? Which tool did you like best?
References
- Lyria 3 Pro: Create longer tracks in more Google products — Google Blog, March 25, 2026
- How developers can use Lyria 3 for AI music generation — Google Blog, March 2026
- Google launches Lyria 3 Pro music generation model — TechCrunch, March 25, 2026
- Lyria 3 — Google DeepMind — Google DeepMind, March 2026
Related Posts:
- AI Music Making: Suno vs Udio 2026 Comparison — The duopoly before Lyria 3 Pro arrived
- Google Search Live Global Expansion + Gemini 3.1 Flash Live — SynthID watermarking in another context