Gemini Gets Musical: Create Custom Tracks with Lyria 3

2/19/2026

The era of personalized AI creativity has hit a high note. On February 18, 2026, Google announced a transformative update to its Gemini app, introducing custom music generation powered by Lyria 3, Google DeepMind’s latest and most advanced generative music model. Moving beyond text and image generation, Gemini now empowers users to become instant composers, turning abstract ideas or visual memories into catchy, high-quality 30-second tracks in a matter of seconds. The Lyria 3 Leap Rolling out today in beta, Lyria 3 represents a significant upgrade in AI audio synthesis. Unlike its predecessors, this model removes the friction of songwriting—users no longer need to provide lyrics, as the AI generates them contextually based on the prompt. It also offers granular control over style, vocals, and tempo, resulting in tracks that are musically complex and startlingly realistic. The feature operates through two intuitive modes: Text to Track: You can describe a specific genre, mood, or memory. For instance, asking for a "fun afrobeat track with a true African vibe about childhood memories and home-cooked plantains" yields a fully produced song that matches that specific atmosphere. From Visuals to Audio: Users can upload a photo or video—like a clip of a dog hiking—and watch Gemini analyze the visual content to compose a track with lyrics that perfectly fit the mood. YouTube Dream Track and Visual Art To make these creations shareable, every track comes with custom cover art generated by the Nano Banana model. But the ecosystem play goes deeper: Creators can now explore Lyria 3 on YouTube’s Dream Track. Already available in the U.S. and now expanding globally, this integration allows YouTubers to generate unique, royalty-free soundtracks or lyrical verses for their Shorts, elevating the production value of short-form video content. Safety and Verification with SynthID In line with Google’s commitment to responsible AI, security is baked into the core of this release. All tracks generated within Gemini are embedded with SynthID, an imperceptible watermark designed to identify content created by Google AI. Furthermore, Gemini’s verification tools have been expanded; users can now upload audio files to check for SynthID markers, adding a layer of transparency to the AI landscape. Google has also implemented strict safeguards regarding copyright. The tool is designed for original expression, not mimicry. Prompts naming specific artists will use them only as broad creative inspiration for style, ensuring outputs do not violate intellectual property rights. Filters are in place to check against existing content, and a reporting mechanism is available for potential violations. Availability Lyria 3 is available starting today on desktop (with mobile rolling out shortly) for users aged 18+ in English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese. While the feature is free to try, Google AI Plus, Pro, and Ultra subscribers will enjoy higher generation limits, ensuring power users have the freedom to experiment fully.