Lyria 3.5 is here. What Ai music in Gemini means for your business

On 4 September 2026, Google made Lyria 3.5 available in the Gemini app. In plain terms: you describe the music you want in a sentence, and Gemini writes and records it, up to three minutes long, free, on the web and in the mobile app.

Music generation is not new in Gemini. Lyria 3 arrived in the app on 18 February 2026, so this week is a better-sounding version rather than a new button. The part the coverage skips is the practical side: every track carries an inaudible watermark, a full-length track needs you to switch the model, and Google publishes no number for how many tracks you get before the limit stops you.

What actually changed?

Google says Lyria 3.5 is “our best-sounding music generation model” and that it “brings more expressive vocals and richer musical arrangements”. Inside the Gemini app you can now pick or describe a genre, choose vocal or instrumental, start from a template, and choose a short or a longer track. Google lists the everyday uses itself: “a custom backing track for a video, a unique brand jingle, or just a personalized ringtone”. The same model is in the Gemini API, Google Flow Music, Google AI Studio and Google Vids.

Source: Google, “Create your best tracks yet with Lyria 3.5 in Gemini”, 4 September 2026

What can you actually make with it?

Short, generic music that sits behind something else. That is the honest description, and for most small businesses it is exactly the kind of music they keep paying for:

• A backing bed for a product, property or site-visit video on Instagram or LinkedIn.
• An intro and outro for a podcast. Google’s own worked example in the Gemini help pages is the prompt “Create an intro track for my podcast about cats.”
• Hold music or an on-hold message bed for the phone system.
• A short jingle for a radio-style social advert.

Tracks download as an MP3 audio file, or as an MP4 video with cover art if you want something to post directly. Through the Gemini API the audio is 44.1 kHz stereo, which is CD sample rate and more than enough for social video or a phone line.

Does it cost anything?

No, not to start. Google says Lyria 3.5 is “available to all users globally on the web and in the mobile app”. Three conditions come with that. You must be 18 or over. You need Gemini Apps activity switched on, which means Google keeps a record of what you generate. And there is a cap: Google’s help pages say “There are limits on how many tracks you can create with Lyria” and that you will get a notification when you pass it, then either upgrade your subscription or wait for the limit to refresh. Google does not publish the number, which is the same pattern as the flexible usage limits it brought to Gemini Notebook. One more detail that catches people out: to get a full track rather than a short one, you have to switch the model to Pro in the Gemini model picker.

Is the music marked as Ai-made?

Yes, always, and you cannot turn it off. Google’s help page says “Each generated track will have an embedded digital watermark to help identify it as AI-generated.” Google DeepMind is more specific: “SynthID embeds a watermark into any audio generated or published through our AI music generation model Lyria or the podcast generation feature of Notebook LM.” The watermark is inaudible, so it does not affect how the track sounds, but it does mean a platform with the right detector can tell how the music was made. That is the same direction of travel as text watermarking in Claude: assume anything generated can be identified as generated.

Can you use a Gemini track in a paid advert?

Google’s music pages do not set out a separate music licence, so the general terms are what apply. Google’s Terms of Service, last modified on 30 July 2026, say “Your content remains yours, which means that you retain any intellectual property rights that you have in your content.” Google’s prohibited use policy for generative Ai bars “Misrepresenting the provenance of generated content by claiming it was created solely by a human, in order to deceive.” Read together: nobody at Google is claiming your track, and nobody is asking you to announce it, as long as you are not passing it off as a human composition to deceive someone. Before a track goes into a paid campaign, check the rules of the platform you are advertising on as well as Google’s, and keep the prompt and the date on file.

What to ignore

Ignore the “Ai has come for musicians” framing that follows every one of these releases. It is not the question in front of a small business and it will not change what you do on Monday. Ignore, too, the urge to sign up for a specialist music tool because this one is only “consumer grade”. If the job is 30 seconds of music behind a video, the tool you already pay nothing for and are already signed in to is the right answer.

What should a small business do next?

Two low-effort steps. First, open gemini.google.com, choose Create music from the add-files menu, and prompt it for the exact thing you last paid for: “Create a calm 30 second instrumental bed for a property walkthrough video, light piano, no vocals.” Switch the model to Pro if you want the full-length version. Second, decide once, in writing, what your business will and will not use Ai-generated music for, so nobody has to ask again. If you already make video in Google’s tools, the same model now sits inside Google Vids, which turns a document into a narrated video summary, so the soundtrack and the video can come from one place. If you would rather someone else worked out where this fits, that is what a Breezi membership is for.