Lyria 3.5 Explained: Gemini's Music Maker Is Free, With Three Catches

Google shipped a new music model into the Gemini app today, and unlike most of what Google ships, this one is free on every plan. On September 4, 2026, Google introduced Lyria 3.5, calling it "our best-sounding music generation model," and said it "is available to all users globally on the web and in the mobile app." That sentence is accurate. It also leaves out the three conditions that decide whether you can actually use it.

Key Takeaways

• It is free, and that is not new. Google's plan comparison ticks "Music generation" for users without a Google AI plan and on AI Plus, Pro and Ultra. Lyria 3.5 is a better model on a feature that was already free.
• You have to be 18. Google's music generation overview says "You must be 18+ to enjoy the experience," and the help article makes it requirement number one. Image generation in the same app starts at 13.
• You have to leave Keep Activity on. Google's Lyria help article lists "Have Keep Activity on" as a requirement to generate music at all. That is the switch that decides whether Google saves your Gemini chats.
• Full-length tracks need the Pro model. Google's instruction is blunt: "To create a full track, switch the model to Pro." Short tracks run about a minute; full ones run about two to three.
• What is actually new is control. Genre selection, a vocal-versus-instrumental switch, a template gallery, longer tracks, and generating a track from an uploaded photo.
• Every track is watermarked. Google says all tracks made in the Gemini app carry SynthID, and that you can now upload an audio file and ask Gemini whether Google AI generated it.
• Paying buys volume, not quality. The plan table gives AI Plus 2x the standard usage limits, AI Pro 4x, and AI Ultra 5x or 20x higher than Pro. The model is the same on all of them.

Do You Get Lyria 3.5? Availability by Plan

Everyone with a Gemini account and an 18th birthday behind them. This is the row from Google's own feature availability table, read on September 4, 2026:

Prices from Google's subscriptions page, re-checked September 4, 2026. Google says limits refresh "every 5 hours until you reach your weekly limit," and warns that media generation "will consume more of your usage" than ordinary chat. So the free plan gets the same Lyria 3.5, and hits the wall sooner.

The Three Catches Google Left Out of the Announcement

The blog post names none of these. All three come from Google's own help pages, and each one can stop you cold.

You must be 18. Both of Google's music surfaces say so independently. This is stricter than the rest of the Gemini app, where you can generate images at 13 and only editing them requires 18. If you are between 13 and 17, music generation is the feature you do not get.

You must leave Keep Activity on. Google's help article lists it as a flat prerequisite: to generate music, "Have Keep Activity on." Keep Activity is the setting that decides whether Google saves the conversations you type into Gemini, and it is the switch privacy-minded readers are most likely to have already turned off. Our guide to turning off Google's AI features explains where it lives and what it does. The trade is now explicit: switch it off and you keep your chats out of your account, but Gemini stops making music.

Full tracks need the Pro model. Google's step-by-step carries an "Important" note saying "To create a full track, switch the model to Pro." Free users can select the Pro model -- Google's plan table marks Gemini 3 Pro as available on every tier, including no plan -- but its own subscriptions page describes the free tier as having "varying access to 3.1 Pro." So a full-length track on the free plan is possible and rationed, not guaranteed.

What Is Actually New in Lyria 3.5

The model got better and the controls got real. Google's list of what changed in the Gemini app is short and specific:

Google also says the model itself brings "more expressive vocals and richer musical arrangements." That is a claim about quality with no published benchmark behind it, so treat it as marketing until you hear it.

How to Make a Track

Google documents the same three steps on computer, Android and iPhone, and the platform-specific help articles differ only in whether you click or tap:

1. Open gemini.google.com or the Gemini app.
2. At the bottom of the text box, choose Upload and tools, then Create music.
3. Optionally upload an image or video for context, then describe the track.

The dropdowns for length, vocals and genre appear at that point. When the track is done, Download offers two formats: MP4 with the cover art, or MP3 audio only. Share produces a link.

Google's prompt advice is worth following because it is specific: lead with genre and era, set tempo and rhythm, name instruments, describe the vocal, and either describe the lyric topic or supply your own text with structure tags like [Verse 1].

What Happens to the Music You Make

Two things Google publishes, and one it does not.

Every track carries SynthID, which Google describes as "our imperceptible watermark for identifying Google AI-generated content." Google has also extended detection to audio: you can upload a file and ask Gemini whether it was generated using Google AI, and it will check for SynthID and reason about the answer. That cuts both ways. It means AI music is identifiable, and it means anything you make is identifiable as AI too.

What Google does not say anywhere on either page is whether you can use the result commercially. There is no licensing statement, no rights grant and no restriction published on the music generation overview or the help article. If you were planning to use a Lyria track as a brand jingle -- one of the uses Google's own announcement suggests -- that question is currently unanswered by Google.

One more limit worth knowing if you use voice: Lyria does not work inside Gemini Live. Google's Live overview says Live "does not support Omni or Lyria," so you have to close the voice screen and go back to typing.

Is It Worth Paying For?

No, not for the music alone.

Stay free if you are making the occasional track for a video, a ringtone or a joke in the group chat. You get the same Lyria 3.5 as everyone else. You will hit a usage wall sooner, and the wall resets in five hours.

Consider Google AI Plus at $4.99 if you keep hitting that wall. Double the standard limits is the cheapest fix, and it also switches on the Nano Banana Pro image redo and video generation, neither of which the free plan gets. That bundle, not the music, is the reason to pay.

Do not upgrade for track quality. Google's plan table ticks music generation identically across all four tiers. There is no premium music model behind the paywall, and nothing in Google's announcement suggests one. For how the paid tiers stack up overall, see which AI chatbot you should pay for.