Google DeepMind Music Model Briefing

Lyria 3: Advanced AI Music Generation for Gemini

Lyria 3 is Google DeepMind's latest and most advanced AI music generation model. Released on February 18, 2026, Lyria 3 in the Gemini app transforms text, image, and video prompts into high-fidelity 30-second tracks with auto-generated lyrics, controllable style and vocals, and built-in SynthID watermarking for responsible AI provenance.

First announced: Feb 18, 2026 Product context: Gemini app Output length: 30-second tracks Safety: SynthID watermarking
Try in Gemini

Official Overview

Lyria 3 is positioned as Google's most advanced music creation model, designed to make high-quality composition more accessible while preserving creative control for users. The system supports multimodal prompting and is optimized for fast creation workflows suitable for short-form content and daily creative expression.

Compared with previous Lyria versions, the reported improvements include stronger musical structure, clearer audio quality, richer arrangement logic, and automatic lyric generation in response to prompt intent.

This page is strictly assembled from the provided reference materials and linked sources. It does not claim unpublished features or timelines.

Lyria 3 Capabilities

1. Multimodal Prompting

Users can create music from text descriptions, uploaded images, and in some workflows video context. This enables fast conversion from narrative mood or visual cues to audio output.

2. Automatic Lyrics and Cover Support

Lyria 3 can generate lyrics automatically from user intent and, in Gemini workflows, can pair tracks with generated cover art assets for a more complete music post package.

3. Higher-Fidelity 30-Second Music Output

Available outputs focus on 30-second tracks with improved coherence and musical section quality, including recognizable intros, verse-like development, transitions, and closings.

4. Creative Controls

Prompt-level control can target genre direction, vocal style, tempo, and mood. Detailed prompting generally improves precision and repeatability of outputs.

Technology, Safety, and Responsible AI

SynthID Audio Watermarking

Reported Gemini-generated Lyria 3 audio includes imperceptible SynthID watermarking for provenance support. This helps identify content generated or edited by Google AI systems.

Verification Flow

Gemini's verification capability can evaluate uploaded audio for SynthID signals and related indicators, supporting trust and transparency use cases.

Style and Rights Safeguards

Reference materials state that Lyria 3 is designed to encourage original expression rather than direct artist imitation. Prompt mentions of specific artists are handled as broad inspiration signals, and reporting pathways exist for suspected rights violations.

Availability and Access Scope

As documented in the references, Lyria 3 is integrated in the Gemini app and available to users aged 18+. Rollout spans web first, with mobile availability following. The model supports multiple languages, including English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese, with additional language support expected over time.

Free usage is available with plan-dependent generation limits. Google AI Plus, Pro, and Ultra tiers are reported to provide higher creation ceilings.

Date context: statements on this page reflect source reporting from February 18-19, 2026.

Market Context and Reported Positioning

In source comparisons against Suno and Udio, Lyria 3 is described as strong in perceived audio quality and tightly integrated user experience through Gemini. At the same time, current 30-second output length is a practical limitation when full-length song production is required.

Enterprise API workflows currently remain centered on Lyria 2 through Vertex AI (`lyria-002`) according to the provided material, while Lyria 3 API availability is described as a watch area rather than a confirmed public endpoint.

Access Matrix (Reference Snapshot)

Platform Model What You Get
Gemini app Lyria 3 30-second tracks, vocals, lyrics, and cover art support in a consumer workflow.
YouTube Dream Track Lyria 3 AI soundtrack generation for short-form video creation.
Vertex AI / Google Cloud Lyria 2 Enterprise API access for instrumental generation and production integration.
Music AI Sandbox Lyria 2 Professional tooling access through waitlist-based environments.
MusicFX DJ / Google Labs Lyria 2 Free-form DJ-style experimentation workflows.

Prompting Guidance (From References)

  • Use detailed prompts with explicit mood, genre, instrumentation, and vocal direction.
  • Try hybrid style prompts (for example, classical + EDM) to test creative boundaries.
  • When supported, add image context to improve emotional and visual alignment.
  • Start broad, then refine with iterative prompt constraints for better control.

Pragmatic guidance: precise prompt intent usually improves output consistency.

Trust, Evidence, and Entity Signals (EEAT-Oriented)

Editorial Identity

Publisher context: reference-based technical communication page for Lyria 3, with explicit source links and date-scoped claims.

Method

This page consolidates the local briefing files in ref/info.md and ref/Google Lyria 3 综合调研报告.md, then anchors key claims to cited public sources.

Primary Entities

Google DeepMind, Lyria 3, Gemini app, SynthID, Vertex AI, YouTube Shorts, and related Lyria family variants.

Sources

  1. Google DeepMind - Lyria model page: https://deepmind.google/models/lyria/
  2. The Keyword (Google Blog, Feb 18, 2026) - Gemini can now create music: https://blog.google/innovation-and-ai/products/gemini-app/lyria-3/
  3. Reddit discussion snapshot (market feedback context): https://www.reddit.com/r/singularity/comments/1r87h60/lyria_3_google_deepminds_music_generator/
  4. Tom's Guide hands-on article (Feb 18, 2026): https://www.tomsguide.com/ai/i-tried-geminis-new-lyria-3-in-app-ai-song-generator-and-it-turned-my-to-do-list-into-a-punk-rock-anthem

External links are marked with rel="nofollow noopener noreferrer" as requested.