Lyria 3 is Google DeepMind's most advanced AI music model. Generate professional 30-second tracks with vocals, lyrics, and cover art from text, images, or video — free inside the Gemini App. Master Lyria3 music creation with expert prompts, tutorials, comparisons, and developer resources.
Lyria 3 is the third-generation music AI model from Google DeepMind, representing the most significant leap in AI music generation to date. It turns any idea — a text prompt, a photograph, or a video clip — into a studio-quality music track complete with vocals, auto-generated lyrics, multi-instrumental arrangements, and AI cover art.
Six powerful capabilities that make Lyria 3 the most versatile AI music generator available today.
Describe your idea in natural language — mood, genre, tempo, instruments, dynamics — and Lyria 3 transforms it into a complete, professional 30-second music track with full arrangement.
Upload any photo — a sunset, a cityscape, a party scene — and Lyria 3 analyzes the visual atmosphere, mood, and context to generate a perfectly matching AI soundtrack automatically.
Feed Lyria 3 a video clip and it creates a custom soundtrack that matches the pacing, emotion, and visual narrative — ideal for content creators who need instant, sync-ready background music.
No need to write your own lyrics. Lyria 3 auto-generates contextual lyrics and sings them with realistic, high-fidelity vocals in 8 languages — including English, Japanese, Korean, and Spanish.
Professional-grade 48kHz stereo output with crystal-clear audio fidelity, layered arrangement behavior, and natural musical flow suitable for commercial projects, podcasts, and social media content.
Every Lyria 3 music track is embedded with Google's imperceptible SynthID audio watermark for responsible AI identification — surviving compression, speed changes, and even re-recording through speakers.
No musical experience required. Open, describe, and generate professional tracks with Lyria3 music inside the Gemini App.
Visit gemini.google.com on desktop or mobile. Select "Create music" from the Tools menu. Free for all users aged 18+ globally where Gemini is available.
Type a detailed text prompt with genre, mood, tempo, vocal style, and lyrics theme. Or upload an image or video and let Lyria 3 create the perfect soundtrack automatically.
In seconds, Lyria 3 generates a 30-second pro-quality track with vocals, lyrics, instrumentals, and AI cover art. Listen, iterate, download, and share with the world.
From pop to phonk, classical to K-pop — Lyria 3 generates authentic, high-quality tracks across every major genre and style fusion.
How Google's Lyria 3 music generator compares to Suno, Udio, Stable Audio, and Meta MusicGen across the features that matter most.
| Feature | Lyria 3 (Gemini) | Suno v4 | Udio | Stable Audio | MusicGen (Meta) |
|---|---|---|---|---|---|
| Track Duration | 30 seconds | Up to 4 minutes | 32s–130s | Up to 3 minutes | ~30 seconds |
| Vocals & Lyrics | ✓ Auto-generated | ✓ Full vocals | ✓ Full vocals | ◯ Limited | ✗ Instrumental only |
| Image to Music | ✓ Native support | ✗ | ✗ | ✗ | ✗ |
| Video to Music | ✓ Native support | ✗ | ✗ | ✗ | ✗ |
| Audio Quality | 48kHz Stereo | 44.1kHz | 44.1kHz | 44.1kHz | 32kHz Mono |
| Languages | 8 languages | Multiple | Multiple | English primarily | English primarily |
| Free Access | ✓ Free in Gemini | ◯ 50 credits/day | ◯ 100 credits/mo | ◯ 3 downloads/mo | ✓ Open source |
| AI Watermark | ✓ SynthID | ✗ | ✗ | ✗ | ✗ |
| Developer API | ◯ Lyria 2 on Vertex AI | ◯ Limited | ◯ Enterprise | ✓ API available | ✓ Open weights |
| Editing Tools | ✗ Not yet | ✓ Virtual DAW | ✓ Extend/Edit | ◯ Basic | ✗ |
Copy-paste these proven Lyria 3 prompts to generate amazing AI music. Each template targets a specific genre, mood, or use case.
From YouTube Shorts to podcast intros, game soundtracks to ad campaigns — Lyria 3 fits seamlessly into every creative workflow.
Generate unique, copyright-free soundtracks for Shorts with Lyria 3 via YouTube's Dream Track feature. Available in the U.S. and expanding globally.
Create professional, branded intros and outros for your podcast in seconds. No licensing fees, no royalty worries — just describe your vibe and generate.
Prototype game music in real-time with the Lyria RealTime API, or generate individual level themes and menu music from detailed text prompts.
Instantly produce custom background music for product demos, social ads, and marketing campaigns without expensive licensing or composer fees.
How Google DeepMind's Lyria music AI model evolved from experimental YouTube tool to the most accessible AI music generator in the world.
First-generation Lyria model launched alongside YouTube Dream Track. Select creators could generate 30-second soundtracks in the style of participating artists like Charlie Puth, Sia, and John Legend. Instrumental focus with licensed voice cloning.
Lyria 2 brought developer access via Vertex AI (model ID: lyria-002). Higher-fidelity instrumental generation at 48kHz WAV, negative prompting, seed reproducibility, and the companion Lyria RealTime model for interactive WebSocket-based music generation.
The most advanced Lyria model launches in the Gemini App for 750M+ users. First public release with full vocal generation, auto-lyrics, multimodal input (text, photos, videos), 8-language support, SynthID watermarking, and YouTube Dream Track global expansion.
Lyria 3's multimodal image-to-music capability is its most distinctive differentiator. Upload any photograph — a sunset, a bustling city, a family portrait, a product shot — and the AI analyzes visual elements including atmosphere, lighting, emotion, and activity to generate a perfectly matching music track.
This feature transforms the creative workflow for photographers, designers, social media managers, and marketing teams who have rich visual content but need audio to match. No music theory, no licensing fees — just upload and generate.
Try Image to Music →A sunset beach photo becomes ambient lo-fi. A party scene becomes upbeat dance music. A snowy mountain becomes cinematic orchestral.
Lyria 3 is the first Lyria model to generate complete songs with automatic lyrics and realistic AI vocals. Specify a theme, emotion, or story — and the model writes contextual lyrics with proper verse-chorus-bridge structure and sings them in your chosen language.
Supported languages: English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese. Use the Lyrics: prefix in your prompt to provide custom lyrics, or describe vocal characteristics like tone, gender, and style for precise control.
Every Lyria 3 music track is embedded with Google DeepMind's SynthID — an imperceptible watermark that enables verification of AI-generated audio content for responsible use and copyright transparency.
SynthID converts audio to a 2D spectrogram, embeds the invisible watermark, then reconstructs the waveform.
The watermark persists through MP3/AAC compression, speed changes, pitch shifting, and even physical re-recording.
Upload any audio file to Gemini and ask if it was generated by Google AI. Over 10 billion pieces of content watermarked.
YouTube's Dream Track — now powered by Lyria 3 — lets creators generate unique AI soundtracks directly within the Shorts creation flow. Previously U.S.-only, Dream Track is expanding worldwide in 2026.
Type a text prompt describing your desired mood, genre, or style, and Dream Track generates a custom background track for your Short. All generated tracks are tagged as AI-created for transparency. YouTube also offers Music Assistant for royalty-free instrumental backing in Creator Music.
Lyria 3 is free for all users 18+ in the Gemini App. Premium subscribers get higher generation limits.
Build interactive music applications with the Lyria RealTime API, or integrate AI music generation into production workflows with Lyria 2 on Vertex AI.
Lyria RealTime enables real-time, interactive music generation through bidirectional WebSocket connections. Generate continuous 2-second audio chunks with adjustable BPM, density, scale, key, tempo, and brightness. Ideal for live DJ applications, interactive art installations, game dynamic soundtracks, and music experimentation tools.
Available via the Gemini API and Google AI Studio. Supports weighted prompts for real-time style blending and smooth musical transitions.
Structure your prompts with these dimensions for maximum control over Lyria 3's output quality and consistency.
Specify primary genre and sub-genre. Blend styles for unique results: "jazz-infused lo-fi" or "cinematic trap."
Include exact BPM (60-180+) or descriptive pace: "slow ballad," "uptempo dance," "moderate groove."
Name specific instruments: "acoustic guitar, upright bass, brushed drums, warm Rhodes piano."
Describe emotional arc: "starts melancholic, builds to hopeful crescendo." Use dynamics: "quiet verse, powerful chorus."
Specify voice type, gender, language, and lyrics theme. Use "Lyrics:" prefix for custom lyrics or "[auto-generate]."
Guide arrangement: "intro, verse, chorus, bridge." Specify transitions: "smooth build," "sudden drop."
Every AI music generator has quirks. Here are the most common Lyria 3 issues and proven workarounds to get better results.
Fix: Be more specific with genre, instruments, and tempo. Remove vague descriptors. Add "only" before key elements: "only acoustic instruments, no electronic sounds."
Fix: Provide custom lyrics using the "Lyrics:" prefix instead of relying on auto-generation. Specify lyric tone: "poetic, metaphorical" or "conversational, casual."
Fix: Reinforce your style throughout the prompt. Add "consistent style throughout" and "maintain [genre] energy from start to finish."
Fix: Generate multiple complementary clips with related prompts and combine them in a DAW. Structure each clip as a distinct section (intro, verse, chorus).
Understanding the rights, risks, and compliance requirements for using AI-generated music from Lyria 3 in commercial projects.
Google does not claim ownership of content you generate with Lyria 3. However, all use is subject to Google's Terms of Service and Generative AI Prohibited Use Policy, which explicitly prohibit generating content that infringes copyrights, misleads audiences, impersonates individuals, or violates others' rights.
Output filters check generated content against existing copyrighted works. All tracks include SynthID watermarks for provenance verification. For enterprise use, Vertex AI offers IP indemnification for Lyria 2 API-generated content.
Key recommendation: Always disclose AI-generated content where required by platform policies, review your specific use case against Google's terms, and never generate content designed to mimic or impersonate specific artists.
How the top AI music tools stack up across quality, features, accessibility, and value. Lyria 3 enters a competitive field led by Suno and Udio.
Best for: Free access, multimodal input, Google ecosystem integration. 750M+ Gemini users, image/video to music, 8 languages.
Best for: Full-length songs, editing tools, virtual DAW. Industry leader with 3+ minute tracks and extensive creative control.
Best for: Audio quality, extend/edit workflow. Backed by Universal Music Group with advanced generation capabilities.
Best for: Clear licensing tiers (Personal/Creator/Enterprise), API access, and transparent commercial use policies.
No AI music generator is perfect. Here are Lyria 3's current limitations and practical solutions for working around them.
Workaround: Generate multiple themed clips and stitch them in a DAW (GarageBand, Audacity, Logic Pro). Structure prompts as "intro," "verse," "chorus" for cohesive transitions. Premium subscribers may get longer limits.
Workaround: Iterate through refined prompts. Generate multiple variations and pick the best. Export to external editors for post-processing, mixing, and mastering.
Workaround: Use third-party AI stem separation tools (like LALAL.ai or Moises) to extract individual tracks from the mixed output for further editing and remixing.
Workaround: Use the Lyria 2 API on Vertex AI for programmatic access (instrumental only). For real-time interactive music, explore Lyria RealTime via the Gemini API and WebSocket connections.
Generate AI music with vocals and lyrics in English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese — with more languages planned.
The AI music generator market was valued at $5.2 billion in 2024 and is projected to reach $60.4 billion by 2034, growing at 27.8% CAGR. Lyria 3's launch accelerates this transformation.
Projected AI music market by 2034
Suno's 2025 funding round valuation
Annual growth rate (CAGR) through 2034
Direct links to Google's official Lyria 3 pages, documentation, and developer tools.
Official model page with technical overview
Official launch post (Feb 18, 2026)
Try Lyria 3 music generation free
Official music generation prompting tips
Developer API for instrumental generation
How Google watermarks AI-generated content
Lyria 3 is Google DeepMind's most advanced AI music generation model, launched February 18, 2026. It generates 30-second, 48kHz stereo music tracks with full vocals, auto-generated lyrics, multi-instrumental arrangements, and AI-generated cover art from text prompts, uploaded images, or videos inside the Gemini App. The model uses advanced audio modeling techniques to produce coherent musical structures with natural verse-chorus-bridge progression. All outputs are embedded with SynthID watermarks for responsible AI content identification.
Yes, Lyria 3 is available free in beta inside the Gemini App for all users aged 18 and above globally where Gemini is available. Free users can generate 30-second music tracks with vocals, lyrics, and cover art. Google AI Plus ($19.99/month), Pro, and Ultra subscribers get higher generation limits and priority access to new features.
Lyria 3 differentiates with multimodal input support (text, image, and video prompts), deep integration with Google's ecosystem (Gemini App and YouTube Dream Track), 8-language vocal generation, SynthID watermarking for content provenance, and free access for 750M+ Gemini users. Suno offers longer tracks (3+ minutes) and a virtual audio workstation for editing. Udio provides extended generation with credit-based pricing and high-quality audio output. Lyria 3's unique image-to-music and video-to-music capabilities are not available in any competitor.
Lyria 3 supports 8 languages at launch: English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese. The model generates both lyrics and vocals in all supported languages, with plans to expand language coverage over time. Quality and genre coverage may vary by language.
Google does not claim ownership of content you generate with Lyria 3. However, commercial use is subject to Google's Terms of Service and Generative AI Prohibited Use Policy, which prohibit content that infringes copyrights, misleads audiences, or violates others' rights. All outputs carry SynthID watermarks. For enterprise use cases requiring IP indemnification, consider the Lyria 2 API on Vertex AI. Always disclose AI-generated content where required by platform policies.
As of February 2026, Lyria 3 does not have a standalone public API. It is accessible through the Gemini App consumer interface and YouTube Dream Track. Developers can use the Lyria 2 API on Vertex AI (model ID: lyria-002) for programmatic instrumental music generation at $0.06 per 30-second clip. The Lyria RealTime API is available via the Gemini API for real-time interactive music generation using bidirectional WebSocket connections. A Lyria 3 API is expected to launch in the future.
SynthID is Google DeepMind's imperceptible digital watermarking technology for AI-generated content. For Lyria 3 audio, SynthID converts the audio waveform into a 2D spectrogram, embeds an invisible watermark pattern, then reconstructs the audio. The watermark is inaudible to humans but detectable by verification tools. It survives MP3/AAC compression, speed changes, pitch shifts, and even physical re-recording through speakers. Over 10 billion pieces of content have been watermarked across all SynthID modalities.
Yes, multimodal input is one of Lyria 3's most distinctive features. Upload any photograph or video clip, and Lyria 3 analyzes the visual content — including atmosphere, lighting, setting, emotion, and activity — to generate a soundtrack that matches the mood and context. A sunset beach photo produces ambient lo-fi, a party scene generates upbeat dance music, and a snowy mountain landscape creates cinematic orchestral scoring. This capability is unique to Lyria 3 among major AI music generators.
Lyria 3 generates 30-second music tracks in the Gemini App for free users. Google AI Plus, Pro, and Ultra subscribers may access higher duration limits (exact limits not yet confirmed). The Lyria 2 API on Vertex AI generates instrumental clips of approximately 32.8 seconds at 48kHz WAV quality. For longer compositions, you can generate multiple complementary 30-second clips and combine them in external audio editing software.
Lyria 3 supports 100+ music genres including pop, hip-hop, lo-fi, K-pop, classical, jazz, EDM, ambient, rock, afrobeat, funk, Motown, phonk, R&B, country, reggaeton, metal, blues, folk, soul, gospel, trap, house, techno, dubstep, drum and bass, synthwave, vaporwave, bossa nova, and many more. You can also blend genres in your prompts for unique hybrid styles like "jazz-infused lo-fi" or "cinematic trap." Quality and authenticity may vary by genre.
Tracking every development in Google's Lyria 3 music generation model and ecosystem.
The comprehensive Lyria 3 resource hub goes live with tutorials, prompt library, comparisons, developer docs, and commercial use guide.
Google DeepMind launches Lyria 3 globally in the Gemini App with text/image/video-to-music generation, auto-lyrics, 8-language vocals, and SynthID watermarking.
YouTube's Dream Track feature, now powered by Lyria 3, begins expanding from U.S.-only to creators worldwide.
Stay ahead with curated prompt templates, new feature announcements, comparison updates, and developer tools delivered to your inbox every week.
Free. No spam. Unsubscribe anytime.
Experience the future of AI music generation. Open the Gemini App, describe your idea, and generate professional-quality music in seconds — free for everyone.