Google officially launched Lyria 3, its most advanced AI music generation model to date. Developed by Google DeepMind, the new model is now integrated directly into the Gemini app (web and mobile), marking a significant leap in how users can interact with and create high-fidelity music.
The launch was timed alongside several major AI announcements at the India AI Impact Summit 2026 in New Delhi, underscoring Google’s commitment to multimodal creative tools.
Key Features of Lyria 3
Lyria 3 moves beyond simple “loop generation” to focus on structured, complex compositions.
- Prompt-to-Song: Create 30-second high-fidelity tracks from text prompts. Unlike previous versions, Lyria 3 can automatically generate lyrics and vocals based on your description.
- Multimodal Inputs: You can upload a photo or video and ask Gemini to “compose a track that fits this mood perfectly.”
- Greater Control: Users can now specify genre, mood, tempo, and even specific vocal styles (e.g., “a nostalgic afrobeat track with a true African vibe”).
- Nano Banana Integration: For every track generated, the Nano Banana model creates custom, high-fidelity album cover art to match the song’s theme.
- SynthID Watermarking: All tracks are embedded with Google’s imperceptible SynthID watermark, allowing anyone to verify if a piece of audio was generated by Google AI.
Capabilities & Availability
| Feature | Detail |
| Duration | 30-second polished tracks (extendable via API). |
| Languages | Supports prompts in English, Hindi, German, Spanish, French, Japanese, Korean, and Portuguese. |
| Access | Free for all users 18+; AI Premium/Ultra subscribers get higher usage limits. |
| Platforms | Gemini web (Desktop), Android, and iOS. |
| YouTube Integration | Powers the expanded Dream Track for Shorts creators globally. |
Creative Guardrails
Google has emphasized a “responsible” approach to music generation through collaboration with the music industry:
- Artist Mimicry: The model is designed for original expression. If a prompt specifies a famous artist, Lyria 3 will treat it as “broad inspiration” for style rather than mimicking the artist’s specific voice or likeness.
- Copyright Filtering: Built-in filters check outputs against existing content to minimize the risk of accidental plagiarism or rights violations.
- Partner Agreements: The model was trained in alignment with licensing deals (such as the 2025 Universal Music Group agreement) ensuring respect for artist protections.
“The goal of these tracks isn’t to create a musical masterpiece, but rather to give you a fun, unique way to express yourself—whether it’s a jingle for a birthday or a ballad from your pet’s point of view.” — Joël Yawili, Senior Product Manager, Gemini App.
