You don’t need a studio, a band, or a single instrument to make original music anymore. Google just dropped Lyria 3.5, its best-sounding music model yet, right inside the free Gemini app. Here’s how to start making tracks today, even if you’ve never touched a keyboard in your life.
What is Lyria 3.5?
Lyria is Google DeepMind’s music generation model. You tell it what you want, and it writes a full track for you: melody, chords, arrangement, even vocals if you ask for them. It’s the same idea as asking ChatGPT to write an email, except the output is actual music instead of text.
The 3.5 update, announced September 4, 2026, is the biggest jump in the family so far. Google says it brings more expressive vocals and richer arrangements, with higher overall fidelity. In plain language: the songs sound fuller and less obviously machine-made than what Lyria could do before.
What makes this one different for normal people is where it lives. Lyria 3.5 isn’t buried in a research demo. It’s in the Gemini app, on the web and mobile, and it’s available to all users globally. No waitlist, no API key, no paid plan required to try it.
Check the official announcement here for the full rundown.
How to make music with Lyria 3.5 in Gemini
The whole thing takes about two minutes the first time. Here’s the flow:
- Open Gemini at gemini.google.com or in the mobile app. You’ll need a Google account, which you almost certainly already have.
- Find the music option. Gemini now offers track creation alongside its text and image tools. The exact menu label varies by device, so look for anything music-related if you don’t see it immediately.
- Describe your track. You can pick a genre directly, like lo-fi, hip-hop, or orchestral, or type a plain-language description. Think “upbeat electronic track for a cooking video,” and it’ll run with that.
- Choose vocal or instrumental. Vocals add a sung or rapped part. Instrumental keeps it clean, which is usually the safer pick for background music.
- Set the length. You can choose short or longer tracks now, which is new. A 15-second clip and a full-length piece are different projects, and it’s nice not to fight the tool for either one.
- Generate and listen. Your track appears in a few seconds. If it’s not quite right, tweak the description or regenerate. Iterating is the whole game, and it’s fast enough that you can do ten versions without losing patience.
One honest note: the controls in Gemini change more often than most apps. If the interface looks a little different from what I described, that’s normal. The core loop, describe, choose style, generate, is what matters.
What you can actually do with the music
This is where it gets fun, because the use cases are real:
- Backing tracks for videos. If you make YouTube videos, TikTok clips, or even slideshow reels, you need music that isn’t a copyright minefield. Generating your own solves the “can I use this?” question.
- Podcast intros and outros. A custom 10-second sting makes your show feel produced. Nobody wants to hear the same free intro every other podcast uses.
- Brand jingles. A tiny business with a jingle used to be a fantasy. Now you can make a forgettable-but-charming one in an afternoon.
- Personal stuff. A birthday track for a friend, a weird ringtone, background music for a family slideshow. Google itself suggests these, and they’re genuinely nice uses.
For content creators, this pairs well with other AI tools. People already use AI video generators like Seedance for the visuals, and MoneyPrinterTurbo to auto-produce social videos. Adding AI music closes the last gap, so the entire video is machine-made, which saves a shocking amount of money on stock subscriptions.
Templates and styles, explained
The template system is the fastest on-ramp. Google ships templates for common situations, like background music or a custom birthday track. You pick one, and Gemini fills in the details for you. It’s the difference between a blank page and a starting point.
Then there are two big style choices:
- Vocal or instrumental. Vocals are impressive when they work and weird when they don’t. Instrumental tracks are hard to mess up, so start there.
- Short or long. Short clips are easier to nail. A 15-second loop only has to be good for 15 seconds. Longer tracks expose more flaws, but they’re also more useful for videos.
Don’t overthink the options. First track, just describe the vibe you want and generate.
Lyria 3.5 vs stock music subscriptions
Before Lyria, your options for non-copyrighted music were stock libraries. Here’s how the two approaches compare:
| Lyria 3.5 in Gemini | Stock music sites | |
|---|---|---|
| Cost | Free with Gemini | Monthly subscription |
| Originality | Unique every time | Same track used by hundreds of videos |
| Speed | Minutes per track | Depends on searching and licensing |
| Control | Describe exactly what you want | Browse what already exists |
| Learning curve | Tiny | Tiny, but licensing rules vary |
Stock libraries still win for one thing: consistency. A human-made track library has predictable quality. AI output can surprise you, in both directions. But for speed, price, and originality, Lyria 3.5 is hard to beat.
The catch
Let’s be honest about the limits, because the hype around AI music is loud.
AI vocals can still sound generated, especially on ballads or songs where the voice carries everything. Lyria 3.5 is better than before, and it’s still not indistinguishable from a human session singer.
Licensing is the other gray area. If you’re monetizing videos, check the current terms before you build your whole channel on AI tracks. Rules have been shifting across every platform, and what’s fine today might not be tomorrow.
And one more thing: don’t use AI music to fake an artist’s style or copy a hit song. That’s how you get a takedown notice and a bad reputation. Make your own thing.
Takeaway
Lyria 3.5 in Gemini is free, fast, and genuinely good for background music, jingles, and personal tracks. Open Gemini today and generate one song. Worst case, you waste two minutes. Best case, your next video finally has music you don’t have to license, search for, or pay for.
And if you’re curious about the rest of Google’s creative AI lineup, the Imagen 4 beginner guide covers image generation the same way. Or, since AI music is everywhere now, learn how to detect AI music on Spotify so you can spot what’s machine-made when you’re just listening.
