By GoogleLearning Objectives
- Understand what Google Flow Music is and how the Lyria model family powers it
- Compare Google's licensed-catalog and provenance posture against Suno and Udio
- Decide which surface fits your work — the consumer app or the Vertex AI developer models
What Is Google Flow Music?
Google Flow Music is Google's AI music generation app: describe a song in plain language and it produces original music with vocals, instrumentation, and structure. It runs on Lyria, the music model family from Google DeepMind, and sits alongside Flow's video and image tools as the audio half of Google's generative-media suite.
The current consumer model is Lyria 3.5, released on July 29, 2026. Google describes four areas of improvement over the prior version: richer and more complex melodic structure, closer adherence to the lyrics you actually write, more realistic and emotionally nuanced vocals with better pronunciation, and finer control over tempo and duration. It arrived as a free upgrade for existing Flow Music users rather than as a new paid tier.
💡Key Concept
Two surfaces, one model family. Flow Music is the consumer app at flowmusic.app where you type a prompt and get a song. Lyria 3 and Lyria 3 Pro are the developer models on Vertex AI, reachable through the API and Media Studio for teams building music generation into their own products. Same underlying family, two very different front doors.
Access
| Surface | Model | Track length | Who it is for |
|---|---|---|---|
| Flow Music app | Lyria 3.5 | Full songs | Creators, marketers, hobbyists |
| Vertex AI | Lyria 3 Pro | Up to three minutes | Developers needing complete compositions |
| Vertex AI | Lyria 3 | Up to thirty seconds | Rapid prototyping, social clips, short-form |
| Google Vids | Lyria | Soundtrack generation | Workspace customers and Google AI subscribers |
The Vertex AI models accept multi-modal input — you can prompt with text or with a reference image — and support user-supplied lyrics, timed lyric generation, instrumental-only output, and structural controls for intros, verses, choruses, and bridges.
Pricing
- All core features at no charge
- Daily AI credit allowance
- SynthID and C2PA on every output
- More credits and generations
- Access to the latest models
- Commercial-use rights for members
- Top up beyond your plan allowance
- Useful for burst projects
- Lyria 3 and Lyria 3 Pro via API
- Media Studio access
- Billed through Google Cloud
⚠️Warning
Confirm current prices at the source. Google publishes Flow Music plan pricing on its own pricing page and folds the paid tiers into the broader Google AI Plus, Pro, and Ultra memberships, which change periodically and vary by country. The structure above is accurate — free core features, paid tiers for more credits and newer models, separate usage-based billing on Vertex AI — but check Google's current numbers before budgeting, and read the terms yourself before relying on the commercial-use grant.
Provenance: The Real Differentiator
Every generative music tool sounds impressive in a demo. What separates them for professional work is what you can prove about the output and what rights you actually hold. This is where Google has taken a visibly different route from its competitors.
Every Lyria output carries two provenance markers. The first is SynthID, Google's inaudible watermark, embedded directly into the audio. It survives the transformations that normally strip metadata — adding noise, MP3 compression, changing playback speed — and it cannot be heard by a listener. The second is C2PA, the Coalition for Content Provenance and Authenticity standard, which attaches tamper-evident credentials describing how the file was made.
The training-data posture is different too. Google states that the Lyria 3 models use material that YouTube and Google have rights to use under their terms of service, partner agreements, and applicable law. Compare that with the position Suno and Udio are defending in court: both face copyright suits from major labels alleging training on copyrighted recordings without permission, and both rely on a fair-use argument that has not yet been tested to judgment.
📝Note
This is a difference in posture, not a guarantee. "Material we have rights to use" is Google's own characterization, and the generative-music rights landscape is unsettled for everyone. Google's Terms of Service and its generative-AI prohibited-use policy still govern what you may do with an output, and both forbid using the tools to violate intellectual-property or privacy rights. If you are putting AI-generated music into advertising, broadcast, or a commercial release, read the current terms and take legal advice — do not rely on a vendor's framing, including this page's summary of it.
Strengths
- Provenance built in: SynthID watermarking plus C2PA credentials on every output, with the watermark surviving compression and speed changes
- A cleaner rights story: Google points to YouTube and Google licensing rather than an untested fair-use defense
- Genuinely free to start: core features carry no charge, with a daily credit allowance and no waitlist
- Two real surfaces: a consumer app for creators and Vertex AI models for teams building music into a product
- Structural control: explicit handling of intros, verses, choruses, and bridges rather than one-shot generation
- Multi-modal prompting: text or a reference image as the starting point
Limitations & Considerations
- Newer to the category: Suno and Udio have longer track records, larger creator communities, and more third-party tutorials and workflows
- Vertex AI pricing is not published in the launch material: budget from the Google Cloud pricing page, not from an estimate
- The standard Lyria 3 model caps at thirty seconds: full compositions need Lyria 3 Pro or the consumer app
- Commercial rights attach to membership, not to the free tier: confirm which plan you are on before using a track commercially
- Tied to the Google account ecosystem: plans interlock with Google AI memberships, which is convenient inside that ecosystem and friction outside it
- The rights landscape is still unsettled industry-wide: a stronger posture is not the same as a resolved question
Best Use Cases
| Task | Why Flow Music |
|---|---|
| Background music for video | SynthID and C2PA make provenance defensible if a platform asks |
| Brand and marketing audio | The licensed-catalog posture suits risk-averse commercial work |
| Short-form social clips | Lyria 3's thirty-second output is built for exactly this |
| Product soundtrack generation | Vertex AI puts the model behind your own interface |
| Songwriting exploration | Structural controls let you iterate on sections rather than whole tracks |
When to choose alternatives:
- Deepest feature set and largest community → Suno AI, the category leader
- Fine-grained clip-level editing and regeneration → Udio, built around extendable clips
- Sound effects and voice rather than music → ElevenLabs
- Cleaning up recorded audio rather than generating it → Adobe Podcast Enhance
Getting Started
- Go to the Flow Music app and sign in with a Google account — core features are free and need no card.
- Start with a plain-language prompt naming genre, mood, instrumentation, and vocal type. Specific beats clever.
- Use the tempo and duration controls rather than re-rolling the whole track when the feel is close but not right.
- Supply your own lyrics when the words matter; Lyria 3.5's improvement here is one of the headline changes.
- Before any commercial use, check which plan you are on and read Google's current terms — the free tier and paid memberships do not carry the same rights.
- For product integration, move to Vertex AI: Lyria 3 Pro for complete compositions, Lyria 3 for short-form.
Key Takeaways
- Google Flow Music is Google's AI song generator, powered by the Lyria model family from Google DeepMind
- Lyria 3.5 shipped July 29, 2026 with better melodic structure, closer lyric adherence, more natural vocals, and finer tempo and duration control — free for existing users
- Developers reach the same family as Lyria 3 Pro (up to three minutes) and Lyria 3 (up to thirty seconds) on Vertex AI
- The real differentiator is provenance: SynthID watermarking plus C2PA credentials on every output, and a training-data posture Google grounds in YouTube and Google rights rather than an untested fair-use defense
- Core features are free; more credits, newer models, and commercial-use rights come with a Google AI membership, and Vertex AI bills by usage
- A stronger rights posture is not a resolved legal question — read the current terms before commercial release
