Google Lyria 3: Features, Access, and What It Can Create

TL;DR: Google Lyria 3 is Google DeepMind's latest generative music model, launched in beta on February 18, 2026. It turns text prompts, photos, and videos into 30-second tracks with vocals, lyrics, and cover art inside the Gemini app (users 18+), and it also powers YouTube Dream Track and Flow Music. This guide covers its verified features, every way to access Lyria 3 AI, how to write effective prompts, and how it compares to Lyria 2 and Lyria 3.5.

What is Google Lyria 3?

Google Lyria 3 is a music generation system developed by Google DeepMind that synthesizes high-quality audio from a text prompt. According to its official model card, it is "Google's most capable music generation model to date," and it marked the third major version of the Lyria family that Google first launched in 2023.

The model takes text as input and outputs both audio (music) and text (lyrics). Technically, it uses latent diffusion applied to temporal audio latents, and it was trained on Google's Tensor Processing Units (TPUs) using JAX and ML Pathways.

Lyria 3 rolled out globally in beta in the Gemini app on February 18, 2026 — available to all users 18 and over in English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese, with plans to expand language coverage. Google AI Plus, Pro, and Ultra subscribers receive higher generation limits.

Key features of Lyria 3

Google's official announcement highlights three ways Lyria 3 improves on earlier Lyria models:

1. Auto-generated lyrics — You no longer need to write your own lyrics; the model generates them from your prompt.
2. More creative control — You can steer elements like musical style, vocals, and tempo.
3. More realistic, musically complex tracks — Outputs are more natural and layered than previous versions.

DeepMind's evaluation of Lyria 3 against Lyria 2 found "significantly" improved audio fidelity, plus better prompt adherence when lyrics are involved — meaning it follows both simple and complex instructions more accurately.

Other confirmed capabilities:

• Text-to-track: Describe a genre, mood, memory, or inside joke and get a track with lyrics or instrumental audio.
• Photo/video-to-track: Upload an image or video and Gemini composes a track that fits its mood.
• Custom cover art: Tracks made in the Gemini app include cover art generated by Nano Banana.
• SynthID watermarking: Every app-generated track carries SynthID, Google's imperceptible AI-content watermark.
• Original-expression guardrails: Lyria 3 is designed not to mimic artists; filters check outputs against existing content.

What can Lyria 3 create?

Google positions Lyria 3 as a tool for personal creative expression rather than professional production: "The goal of these tracks isn't to create a musical masterpiece, but rather to give you a fun, unique way to express yourself."

Real, confirmed use cases from the launch:

• Custom songs for people: A personalized afrobeat track for a family member, complete with lyrics.
• Soundtracks from memories: Turn vacation photos or a video of your dog on a hike into a matching song.
• Fun gifts: Custom birthday cards and novelty tracks like "a ballad from your pet's point of view."
• Shorts soundtracks: YouTube Dream Track, which was first available in the U.S. and is rolling out to more countries, uses Lyria 3 to generate lyrics and backing tracks for Shorts.
• Workplace video audio: Lyria 3 is distributed through Google Vids and Google Cloud/Vertex AI, alongside the Gemini app, Google AI Studio, and Music AI Sandbox.

How to access Lyria 3

Lyria 3 is distributed through these official channels:

• Gemini app (gemini.google.com/music) — the primary consumer entry point, launched February 2026
• YouTube Dream Track — AI soundtracks for Shorts
• Google Flow Music / Producer AI — Google's music-focused studio product
• Google AI Studio and Gemini API — for developers
• Google Cloud / Vertex AI — for enterprise deployments
• Google Vids — AI-assisted video creation for work

Practical steps to try Lyria 3 AI in the Gemini app:

1. Open the Gemini app at gemini.google.com (desktop rolled out first; mobile followed over the following days).
2. Describe an idea in natural language, for example: "a comical R&B slow jam about a sock finding their match."
3. Or upload a photo or video and ask Gemini to use it as inspiration for a track.
4. Wait a few seconds — Gemini returns a 30-second track with lyrics and cover art.
5. Download the track or share it via a share link. Tracks are watermarked with SynthID.

How to write better Lyria 3 prompts

DeepMind's official Lyria prompt guide recommends covering these elements:

• Genre and era: Name a genre or blend them, like "early 90s hip-hop" or "K-pop with a Motown edge."
• Tempo: Request a slow ballad, modern pop, or drum and bass — or an approximate BPM.
• Instruments: Add unexpected instruments ("80s synth over 1950s jazz") or let Lyria choose to suit the genre.
• Dynamics: Describe how sections flow — a quiet piano build into an explosive chorus, for example.
• Vocals: Define a vocal profile (male/female, range, tone) and language.
• Own lyrics: Prefix your lines with "Lyrics:" to supply custom words; add backing-vocal echoes in parentheses.

Keep it simple if you're starting out — a basic prompt like "an upbeat birthday tune" is enough. Detail is a dial, not a requirement.

Lyria 3 vs Lyria 2 vs Lyria 3.5

If you're comparing Lyria 3 AI to the newer Lyria 3.5, the practical difference is that 3.5 targets longer, studio-style generation in Flow Music, while Lyria 3 remains the model inside the Gemini app for quick, shareable 30-second tracks. Both are DeepMind models, and both are watermarked with SynthID.

Safety, limitations, and responsible use

Google states it trained Lyria 3 "very mindful of copyright and partner agreements" and in collaboration with the music community. Confirmed safeguards include output filters against existing content, SynthID watermarking on all app generations, audio verification (upload a file and ask Gemini if it was generated by Google AI), and reporting tools for content that may violate rights.

Known limitations: The Gemini app currently produces 30-second clips — not full songs. And as DeepMind's own model page notes, Lyria "is still working on improving key capabilities," so always check that generated tracks fit your vision. Don't expect artist-faithful mimicry; that's a design choice, not a bug.

Want hands-on prompt templates, walkthroughs, and the latest Lyria 3 updates in one place? Visit the Lyria 3 Pro hub — it's the fastest way to move from "curious about Lyria 3 AI" to making tracks you actually use.

FAQ

Is Google Lyria 3 free? Lyria 3 in the Gemini app is available to all users 18 and over, and Google AI Plus, Pro, and Ultra subscribers get higher generation limits, per Google's announcement. Google has not published a separate Lyria 3 pricing page.

Is Lyria 3 the same as Lyria 3.5? No. Lyria 3 launched in the Gemini app in February 2026; Lyria 3.5 is the newer model released in Flow Music in July 2026 with improved musicality, lyrics, and vocals.

Can Lyria 3 copy real artists? No. Google designed it for original expression and says prompts naming specific artists are treated as broad creative inspiration, with filters checking outputs against existing content.

Does Lyria 3 generate lyrics? Yes — auto-generated lyrics are one of the model's three headline improvements over previous Lyria models.

Sources

• Google DeepMind — Lyria 3 Model Card
• Google DeepMind — Lyria model page
• Google — A new way to express yourself: Gemini can now create music
• Google DeepMind — How to create effective prompts with Lyria
• Music Business Worldwide — Google just launched Lyria 3 — its 'most advanced' AI music generator yet — in the Gemini app

All facts in this article were verified against the official Google DeepMind model card, product pages, prompt guide, and Google's launch announcement as of August 14, 2026. Where Google has not published specifics (for example, pricing), we omit rather than speculate.