You saw a demo of someone typing a few words into the Gemini app and getting back a real song, and now you want to do the same thing — but the moment you open Gemini, there is no obvious "make music" button, no clear pricing, and no explanation of whether you are getting a full track or a short clip. Search results do not help much either: half of them talk about a Lyria 3 Pro API you never asked for, the other half assume you already pay for Google's top tier.
Here is the thing most articles gloss over: "Gemini Lyria 3" is not a separate product you sign up for. It is Google's Lyria music model working inside the Gemini app — and what you actually get depends on which Gemini plan you are on and which Lyria variant your prompt is routed to. Get that mapping right and the confusion disappears.
This guide clears it up in plain terms: what Gemini Lyria 3 is, what you get on each plan, how to prompt it, how images-to-music and the SynthID watermark work, and how the in-app experience differs from the API and Vertex AI paths — plus honest, cheaper ways to run the same model if you do not have a paid Gemini subscription.
What "Gemini Lyria 3" Actually Means
Lyria 3 is Google DeepMind's music generation model. The Gemini app is Google's consumer AI assistant. "Gemini Lyria 3" is simply the first of those running inside the second — you describe the music you want in a normal Gemini conversation, and Lyria generates the audio.
The base Lyria 3 model launched in the Gemini app on February 18, 2026, producing 30-second clips. About a month later, on March 25, 2026, Google announced Lyria 3 Pro, which generates original songs up to 3 minutes long — with intros, verses, choruses, and bridges — including AI-generated vocals and lyrics. So when people say "the Gemini music generator," they could mean either variant, which is exactly why expectations get mismatched.
Two facts worth internalizing before you start:
- The Gemini app is only one of six ways to reach Lyria. The others are Google Vids, Google AI Studio / the Gemini API, Vertex AI, ProducerAI, and pay-as-you-go credit studios like this one. Same underlying model, very different access rules and costs.
- What you get in the app is gated by your plan. A free Gemini account is not the path to full three-minute songs.
What You Get: Lyria 3 Clips vs Lyria 3 Pro Songs
The single biggest source of disappointment is expecting a finished song and receiving a 30-second sketch. This table is the mental model to keep:
| Lyria 3 (base) | Lyria 3 Pro | |
|---|---|---|
| Length | Up to ~30 seconds | Up to 3 minutes |
| Structure | Basic; one idea | Intro / verse / chorus / bridge |
| Vocals & lyrics | Limited | AI vocals + lyrics, 8 languages |
| Best for | Sketches, social clips, testing a prompt | Finished, shareable songs |
| Rule of thumb | Prototype here | Produce here |
A good habit: treat base Lyria 3 as your scratchpad. Nail the vibe with a cheap 30-second clip first, then commit to a full Pro generation once the prompt is right. That one step saves you from burning a longer generation on a direction you were never going to keep.
Which Gemini Plan Do You Need?
This is the question the demos never answer. Access to Lyria inside the Gemini app is a paid Gemini subscriber feature — a free Gemini account does not unlock full music generation, and the higher your tier, the more you can generate.
A few honest caveats, because Google does not publish one tidy chart for this:
- Google has not released a single unified consumer price for Lyria 3 Pro music generation. Tiers, daily track limits, and which plan unlocks Pro-length songs can change, so treat any dollar figure or daily-track number you see online as per third-party reports — verify on Google's current pricing page first.
- Access can also arrive through Google Vids (for Google Workspace customers and Google AI Pro & Ultra subscribers) if your goal is scoring a video rather than exporting a standalone track.
If you are only music-curious, paying for a top-tier plan purely to test Lyria is a lot of commitment — which is where the credit route below is worth a look.
How to Prompt Gemini Lyria 3
Inside the app you are just chatting, but the model responds far better to a structured brief than to "make me a song." Google publishes an official prompt formula, and it works the same way in Gemini as it does through the API:
[Genre & style] + [Mood] + [Instrumentation] + [Tempo & rhythm] + [Vocal style & language] + [Lyrics]A worked example you can adapt:
Dream-pop, nostalgic and warm, layered synths with a soft
electric guitar, mid-tempo around 90 BPM, breathy female
vocals in English.
Lyrics: late-night drives and city lights fading behind us.Three flags do most of the heavy lifting:
- End the prompt with
Instrumental.to get no vocals at all. - Prefix your own words with
Lyrics:so the model sings exactly what you wrote instead of inventing lyrics. - Describe multiple singers ("a male and female duet trading lines") when you want more than one voice.
For full songs on Lyria 3 Pro, the underused trick is the timestamp technique: you script the arrangement with cues like [00:00] intro, [00:30] verse, [01:00] chorus, through to [03:00], so the model changes sections on your schedule instead of looping one texture for three minutes. If your Gemini output feels repetitive, this is usually the fix. There is a full breakdown in our Lyria 3 Pro prompt guide.
Images-to-Music and Other Multimodal Inputs
One genuinely distinctive feature is that Lyria accepts more than text. Both variants take text prompts, your own lyrics, reference images (up to 10), and PDF as inputs. Feed it a moody photograph or a piece of cover art and the model uses it to shape the mood and palette of the track — a fast way to get a coherent vibe without writing a paragraph of adjectives.
What Lyria does not accept is just as important: no audio upload and no voice cloning. You cannot hum a melody or upload a reference song for it to imitate, and you cannot clone a real person's voice. If those are core to your workflow, Gemini Lyria is not the right tool for that part of the job.
What Happens to Your Track: SynthID and Content Credentials
Every clip or song you generate carries an invisible SynthID watermark — Google's system for marking audio as AI-generated — and outputs support C2PA content credentials. This is not something you toggle; it is on by default. It matters for two practical reasons: platforms and rights systems can detect that the audio was AI-made, and if you plan any commercial use you should check the current terms on whatever access path you used, since those terms differ between the consumer app and the enterprise Vertex AI path.
On training data, Google states the model was trained on licensed and permissible data, and that outputs are checked against existing content.
Gemini App vs the API and Vertex AI Paths
If your search led you to talk of endpoints and pay-per-generation billing, that is a different door to the same model. Quick orientation:
| Path | Who it's for | How it's billed |
|---|---|---|
| Gemini app | Consumers who want to chat and create | Paid Gemini subscription |
| Google AI Studio / Gemini API | Developers building apps | Metered, pay-per-generation (preview) |
| Vertex AI | Enterprises, Media Studio users | Metered via Google Cloud (public preview) |
| Credit studio | Anyone wanting Pro songs without a Google plan | Pay-as-you-go credits |
The Gemini app is the least technical option — no code, no Google Cloud project — but it is also the least flexible, and it is subscription-gated. The API and Vertex AI give you programmatic control and metered billing but require developer setup. For a deeper comparison of the developer surfaces, see our note on the Lyria 3 model and its access paths.
No Paid Gemini Plan? The Cheaper Way to Try Lyria 3 Pro
Here is the honest part. If you do not already pay for a qualifying Gemini tier, subscribing just to test music generation is a poor trade — you are committing to a monthly bill for a feature you have not tried.
A pay-as-you-go credit studio runs the same Lyria 3 Pro model without any Google subscription or Cloud setup. You buy credits, generate full three-minute songs when you actually need them, and pay nothing in the months you do not. That is the model behind our AI music generator: the structural control and 48 kHz-class output of Lyria 3 Pro, on a no-subscription basis. You can see exact per-generation credit costs on the pricing page rather than guessing at a monthly tier.
Rule of thumb: if you already live inside Google's paid ecosystem, use Gemini. If you just want full Lyria 3 Pro songs on demand without a recurring bill, use a credit studio.
How to Tell It Worked
A quick, low-friction check before you commit real effort: generate one short base-Lyria-3 clip and listen for three things — does the genre match, does the mood match, and does a vocal come in where you expected (or stay absent if you added Instrumental.)? If all three land in a 30-second sketch, your prompt is ready for a full Pro generation. If one is off, fix the prompt slot responsible first.
Frequently Asked Questions
Is Gemini Lyria 3 free? Music generation inside the Gemini app is a paid-subscriber feature, so the full experience is not free. A free Gemini account will not unlock full song generation. If you want to try Lyria 3 Pro without a subscription, a credit studio with starter credits or a pay-as-you-go model is the cheaper entry point.
Does the Gemini app make full songs or just clips? Both, depending on the variant your prompt reaches. Base Lyria 3 produces clips up to about 30 seconds; Lyria 3 Pro produces songs up to 3 minutes with proper structure. Higher plans are what unlock the longer, Pro-length output.
Can I upload my own song or clone a voice in Gemini? No. Lyria accepts text, lyrics, images (up to 10), and PDF, but it does not support audio upload or voice cloning on any access path.
What languages can the vocals sing in? Eight: English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese.
Is AI music from Gemini watermarked? Yes. Every output carries an invisible SynthID watermark and supports C2PA content credentials by default — you cannot turn it off.
How is this different from the Lyria API? Same model, different front door. The Gemini app is a no-code, subscription-based chat experience; the Gemini API and Vertex AI are metered, pay-per-generation developer paths that require setup. A credit studio sits in between: full Pro songs, no code, no subscription.
The Bottom Line
"Gemini Lyria 3" is just Google's Lyria music model living inside the Gemini app. What you get — a 30-second sketch or a full three-minute song — comes down to which variant your prompt reaches and which paid plan you are on. Prompt it with the official formula, lean on the timestamp technique for structure, and remember every track is SynthID-watermarked.
If you already pay for Google's ecosystem, the Gemini app is a convenient place to create. If you do not, you do not have to subscribe to get Lyria 3 Pro — generate full songs on demand with our credit-based Lyria 3 Pro studio and check the pricing page for exact costs.
