Google · Advanced AI Music Generation Model

Google Lyria 3

Three Major Improvements

Google DeepMind's most advanced music generation model yet. Create 30-second high-fidelity tracks from text or images — with automatic lyrics, realistic vocals across 8 languages, and professional-grade audio that flows naturally from start to finish.

30s
High-fidelity track length
8
Vocal languages
Auto
Lyrics generation
SynthID
Watermarked output
What's new

Three Major Improvements

Launched in beta on February 18, 2026 — integrated into the Gemini app and YouTube Dream Track

01

Automatic Lyrics

No need to provide your own lyrics! They'll be generated for you based on your prompt. Lyria 3 creates lyrics that perfectly match your music's mood, genre, and narrative structure.

02

Creative Control

You have more creative control over elements like style, vocals, and tempo. Define realistic vocal styles and acoustic preferences to find exactly the sound you're looking for.

03

Realistic & Complex

Create more realistic and musically complex tracks with high-fidelity audio that flows naturally from start to finish. Professional-grade quality ready for your projects.

Overview

Model at a glance

Google Lyria 3, developed by Google DeepMind, represents the most advanced music generation model yet. Launched in beta on February 18, 2026, and integrated into the Gemini app, Lyria 3 empowers anyone to create 30-second high-fidelity tracks using text prompts or images. Running on top of Gemini's language model, the AI actually understands narrative structure and writes verses and choruses that work together cohesively.

Lyria 3 improves on previous Lyria models in three critical ways: automatic lyrics generation eliminates the need to provide your own lyrics, enhanced creative control over style, vocals, and tempo gives you precise direction over your sound, and improved audio quality produces more realistic and musically complex tracks. The model supports 8 languages (English, German, Spanish, French, Hindi, Japanese, Korean, Portuguese) and spans genres from pop to funk to Motown. All generated tracks are embedded with SynthID, Google's imperceptible watermark for identifying AI-generated content, and come with custom cover art generated by Nano Banana.

Developer
Google DeepMind
Status
Beta — February 18, 2026
Release Date
February 18, 2026
Architecture
Runs on Gemini LLM
Track Length
30 seconds
Audio Quality
High-fidelity, professional-grade
Languages
8 (EN, DE, ES, FR, HI, JA, KO, PT)
Integration
Gemini app, YouTube Dream Track
Capabilities

Key features

Text to Track

Describe a specific genre, mood, inside joke, or memory to create unique tracks with lyrics or instrumental audio that fits your vibe. Simply type what you want to hear, and Lyria 3 will compose it for you with automatic lyrics, vocals, and instrumentals perfectly matched to your description.

Image/Video to Track

Upload a photo or video and watch Gemini use the content to compose a track with lyrics that fit the mood perfectly. Lyria 3 transforms visual content into custom high-fidelity music, creating soundtracks that capture the essence and emotion of your images.

Automatic Lyrics Generation

No need to write your own lyrics! Lyria 3 automatically generates lyrics based on your prompt that perfectly match the music's mood, genre, and narrative structure. The AI understands storytelling and creates verses and choruses that work together cohesively.

8 Languages Supported

Generate vocals in 8 different languages: English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese. Create music that speaks to global audiences with native-quality pronunciation and natural-sounding vocals in each supported language.

Multiple Genres

Create music spanning genres from pop to funk to Motown, R&B, Afrobeat, and more. Explore global soundscapes and find musical inspiration across diverse styles. Each genre is rendered with authentic instrumentation and production characteristics.

Realistic Vocal Styles

Define realistic vocal styles and direct the details to find the sound you're looking for. Compose with freedom and technical control over vocal characteristics, from smooth R&B to energetic pop, ensuring your track sounds exactly how you envision it.

30-Second Tracks

Generate complete 30-second tracks in seconds. Perfect length for social media content, YouTube Shorts, quick expressions, and sharing with friends. Each track flows naturally from start to finish with cohesive musical structure.

SynthID Watermarking

All tracks are embedded with SynthID, Google's imperceptible watermark for identifying AI-generated content. Ensures transparency and allows verification of whether music was created using Google AI, supporting responsible AI usage.

Auto Cover Art

Each track comes with custom cover art automatically generated by Nano Banana. Makes it easy to quickly share your creations with friends by downloading or simply clicking the share link. Complete visual and audio package ready to go.

Specs

Technical specifications

Model Type

TypeAI Music Generation
DeveloperGoogle DeepMind
ReleaseFeb 18, 2026 (Beta)
ArchitectureRuns on Gemini LLM
IntegrationYouTube Dream Track

Audio Output

Track Length30 seconds
QualityHigh-fidelity, professional-grade
FlowNatural note-to-note progression
ComplexityRealistic & musically complex
SpeedGenerated in seconds

Input Methods

Text PromptsGenre, mood, theme
ImagesPhoto-to-music transformation
VideosVideo content to soundtrack
LyricsAuto-generated (optional manual)
ControlStyle, vocals, tempo

Languages & Genres

Languages8 (EN, DE, ES, FR, HI, JA, KO, PT)
PopContemporary pop music
FunkGroovy funk tracks
MotownClassic Motown sound
MoreR&B, Afrobeat, and global genres

Safety & Verification

WatermarkSynthID (imperceptible)
VerificationAI content detection
CopyrightNo artist mimicking
FilteringContent safety checks
Age18+ users only

Additional Features

Cover ArtAuto-generated by Nano Banana
SharingDownload or share link
UnderstandingNarrative structure
LyricsVerses & choruses cohesively
FormatCrisp, clear tracks
In practice

Use cases

Personal Expression & Fun

Create unique tracks about inside jokes, memories, or special moments. Express yourself in a fun, creative way with custom soundtracks for birthdays, celebrations, or just for laughs. A unique way to express yourself through music.

YouTube Shorts Soundtracks

Enhance your YouTube Shorts with custom soundtracks via Dream Track. Create lyrical verses or vibey backing tracks that take your Shorts to the next level. Perfect 30-second length for social media content with professional-quality audio.

Social Media Content

Generate custom soundtracks for Instagram stories, TikTok videos, and other social media posts. Upload photos or videos and let Lyria 3 create music that perfectly captures the mood. Easy sharing with download or share link options.

Creative Exploration

Explore new genres, play with soundscapes, and find musical inspiration. Experiment with different styles, moods, and vocal characteristics. Perfect for musicians, content creators, and anyone looking to discover new musical possibilities.

Memory & Nostalgia

Create tracks about cherished memories, family moments, or nostalgic experiences. Turn photos from special occasions into custom soundtracks. Perfect for gifts, tributes, or preserving memories in a unique musical format.

Video Background Music

Generate custom background music for personal videos, presentations, or projects. Upload video content and let Lyria 3 create a soundtrack that matches the visual mood. Professional-grade audio ready for your creative projects.

Honest look

Current limitations

30-Second Limit

Tracks are currently limited to 30 seconds in length. While perfect for social media and quick expressions, longer compositions are not yet supported. The focus is on creating fun, shareable moments rather than full-length songs.

Beta Status

Lyria 3 is currently in beta release. Features, quality, and capabilities may continue to evolve based on user feedback and ongoing development. Some outputs may vary in quality as the model continues to improve.

No Artist Mimicking

Designed for original expression, not for mimicking existing artists. If you name a specific artist, Lyria 3 will take this as broad creative inspiration and create a track with similar style or mood, not an exact replica.

Not for Masterpieces

The goal isn't to create musical masterpieces, but rather to provide a fun, unique way to express yourself. Best suited for casual creation, social sharing, and creative exploration rather than professional music production.

Age Restriction

Available only to users 18 years and older. This age restriction ensures responsible use and compliance with content policies. Users must adhere to Terms of Service and Gen AI prohibited use policies.

Language Coverage

Currently supports 8 languages with plans to expand quality and coverage of more languages. Some languages may have better quality than others as the model continues to improve global language support.

FAQ

Frequently asked questions

What is Google Lyria 3?
Google Lyria 3 is Google DeepMind's most advanced music generation model yet, launched in beta on February 18, 2026. It empowers anyone to create 30-second high-fidelity tracks using text prompts or images. Running on Gemini's language model, Lyria 3 understands narrative structure and automatically generates lyrics, vocals, and instrumentals that work together cohesively.
What are the three major improvements in Lyria 3?
Lyria 3 improves on previous Lyria models in three critical ways: (1) Automatic lyrics generation - no need to provide your own lyrics, they're generated based on your prompt; (2) More creative control over elements like style, vocals, and tempo; (3) More realistic and musically complex tracks with high-fidelity audio that flows naturally from start to finish.
How does text-to-track work?
Simply describe a specific genre, mood, inside joke, or memory, and Lyria 3 will create a unique track with lyrics or instrumental audio that fits your vibe. For example, you could say "Create a fun afrobeat track about my mother and memories of her home cooked plantains" and Lyria 3 will compose it with automatic lyrics, vocals, and instrumentals perfectly matched to your description.
Can I create music from images or videos?
Yes! Upload a photo or video and Lyria 3 will use the content to compose a track with lyrics that fit the mood perfectly. For example, upload photos of your dog on a hike and Lyria 3 will create a custom soundtrack that captures the essence and emotion of those images. This image-to-music capability is one of Lyria 3's unique features.
What languages and genres are supported?
Lyria 3 supports 8 languages: English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese, with plans to expand coverage. It can create music spanning genres from pop to funk to Motown, R&B, Afrobeat, and more. Each genre is rendered with authentic instrumentation and production characteristics.
Do I need to write my own lyrics?
No! One of Lyria 3's major improvements is automatic lyrics generation. Lyrics are automatically generated based on your prompt and perfectly match the music's mood, genre, and narrative structure. The AI understands storytelling and creates verses and choruses that work together cohesively. You can also provide your own lyrics if you prefer.
How long are the generated tracks?
Lyria 3 generates 30-second tracks, which is the perfect length for social media content, YouTube Shorts, quick expressions, and sharing with friends. Each track flows naturally from start to finish with cohesive musical structure. The tracks are generated in seconds and come with custom cover art by Nano Banana.
What is SynthID watermarking?
All tracks generated by Lyria 3 are embedded with SynthID, Google's imperceptible watermark for identifying AI-generated content. This ensures transparency and allows verification of whether music was created using Google AI. You can upload any audio file and ask if it was generated using Google AI, and it will check for SynthID.
Can it mimic existing artists?
No. Lyria 3 is designed for original expression, not for mimicking existing artists. If your prompt names a specific artist, Lyria 3 will take this as broad creative inspiration and create a track that shares a similar style or mood, not an exact replica. Filters are in place to check outputs against existing content to protect copyright.
How can I access Google Lyria 3?
You can access Google Lyria 3 directly through SharkFoto. Simply visit SharkFoto.com, select Lyria 3 from the available AI music models, and start creating. SharkFoto provides seamless access to all Lyria 3 features including text-to-track, image-to-music, automatic lyrics generation, 8 languages, multiple genres, and professional-grade audio output.

Ready to Create Your Own Music?

Experience Google DeepMind's most advanced music generation model. Create 30-second tracks from text or images with automatic lyrics, 8 languages, and professional audio.

Try Google Lyria 3 now