Introducing Lyria 3, our new music generation model in Gemini that lets you turn any idea, photo, or video into a high-fidelity track with custom lyrics. From funny jingles to lo-fi beats, you can create custom 30-second soundtracks for any moment. See how it works. 🧵

To try Lyria 3, head to and select “Create music” in the tools menu, then describe what you want to hear. Gemini will transform your prompt into a high-quality track. You can even ask Gemini to take inspiration from a photo or video you upload. Here are three ways to get started 👇

1) Text to Tracks: Describe a specific genre, mood, or topic to create unique tracks with lyrics or instrumental audio that fits your vibe. For example, you could prompt, “Create a 90's skate punk rock track to tell my roommate Ryan to wash the dishes, high energy, fast drums.”

2) Image or Video to Tracks: Upload a photo or video and watch Gemini analyze the content to compose a track with lyrics that fit the vibe. For example, you could prompt, “Use these photos to create a track about my dog Duncan on hikes in the woods.”

3) Templates: Need some more inspiration? Remix a template to create a custom track instantly, or use dynamic suggestions to help you find the exact words to describe the music in your head. Templates and dynamic suggestions are available now on desktop and are rolling out to mobile soon.

All tracks generated in Gemini are embedded with SynthID, our imperceptible watermark for identifying Google AI-generated content. We are also giving you more tools to help identify AI content, broadening our verification capabilities to include audio. Simply upload a file and ask if it was generated using Google AI, and Gemini will check for SynthID.

Lyria 3 is rolling out in beta on desktop today and to the Gemini mobile app over the next several days for all users globally. Try it out by selecting "Create music" in the tools menu. We can’t wait to hear what you make. Share your tracks with us below ↓

🎶

Meme makers when Google gives them Lyria 3 AI tech.

Not bad 🚬🥃

Here comes, the SUNO killer... 🫡

is region locked?, i dont have it, from argentina.

Google is playing the long game here. Free music generation gets people deeper into the Gemini ecosystem.

Could you clarify what audio data your models were trained on? I assume everything was licensed properly and used with full permission from copyright holders, right? Yeah? Sure?

I've got an issue where my entire all my chats disappeared from the sidebar, but queries are still visible in Gemini Activity part. Is there a fix or a known bug that's gonna get fixed

I've scrapped video projects because I couldn't find the right music and didn't want to deal with licensing. A tool that generates custom soundtracks from the video itself removes the last friction point in my content workflow completely.

Just made this awesome sound with @GeminiApp xD

This is a song I made about my day.

Is it not rolled out for everyone or any regional constraints are there? Because I cannot see it in my Gemini. I am Gemini Pro member.

Trained on what?

made a song asking my nephew to study Math :) @GeminiApp named the song "Algebraic Ascension" and its pretty amazing! He loves it 😍 But you gotta get your numbers right :)

Sounds great, thanks for giving me access early.

just in case you are wondering why Google called it "Lyria":

This is so awesome, but I know there's an industry somewhere that's gonna be really upset by this...

Here a few of my tests. Hope you enjoy:

My theme

Guys, this is ... I don't have a proper adjective for it! Thank you!!

Algorithmic composition democratizes sonic branding. Expect a proliferation of ephemeral audio content, devaluing originality while amplifying personalized noise. ⚙️

Ultra user here. Where is the feature?

please allow for audio reference track (i.e. hum a melody, get a track back) i know you all are capable of it. Suno is the only good product that has that, but they gatekeep their API 🥲

Assuming this is still rolling out

The 'Ryan wash the dishes' punk track example is the petty energy I didn't know I needed from AI. 🎸🧼 We are entering the golden age of Passive-Aggressive Generative Media. Why send a text when you can drop a fully produced 90s skate punk anthem about your roommate's hygiene? This is the killer app. 📉

Lyria @GeminiApp Prompt: Melancholic Folk, sound of water dripping in a cave, acoustic guitar with slight fret buzz (realistic), slow tempo, deep emotional vocals, natural reverb, haunting.

And later, would it be possible to create full tracks? And how about generating tracks from I’d like if I had an idea and don’t have instrument to play it but I’m humming it

It's Chinese New Year and I'm about to leave home. So I asked @GeminiApp to compose a farewell track — folk-pop, shifting from festive warmth to quiet solitude. Seconds later, Lyria 3 delivered a 30-second piece that genuinely moved me. 30 seconds is short. But when AI captures emotion this well, it hits different. 🎵

I am generally happy with new releases... this is soo shit , feels like 2010 called.

This is genuinely exciting. Turning everyday moments or random photos into personalized high-fidelity tracks opens up so much creative fun. Looking forward to experimenting with it soon. Great work bringing Lyria 3 to life.

Music producers are sweating. The day you can generate a track faster than you can describe it is the day the industry realizes it's not about the tools anymore - it's about the creative direction. Lyria looks seriously impressive though.

Too much AI man

Turning ideas into full tracks in seconds is wild. We went from typing prompts to basically producing soundtracks for our thoughts. Curious how creators are going to use this.