Discover Google Lyria 3, the new AI music generation model in the Gemini app that creates 30-second tracks with lyrics and custom art.
What Lyria 3 brings into the Gemini app
Google’s Lyria 3 is a generative AI music model built into the Gemini app. From a short prompt, it can produce a 30-second track with lyrics and matching custom art. That combination—audio, words, and a visual—turns a single request into a small package you can listen to, share, or use as a starting point for something longer.
Thirty seconds is short enough to try ideas quickly and long enough to hear structure: verse or hook feel, rhythm, and whether the lyrics fit the mood. You are not getting a full song out of the box; you are getting a focused sketch that is easy to evaluate and iterate on.
How to write prompts that produce usable clips
Treat the prompt like a brief for a session musician, not a one-word genre tag. State the style, tempo feel, emotional tone, and who the track is for. If you need lyrics, say what the song is about and the point of view (first person, storytelling, product demo, comedy). Mention instruments or textures only when they matter—sparse piano versus dense electronic beats changes the result more than vague adjectives.
After the first clip, refine one variable at a time: tighter lyric theme, different energy level, clearer structure (intro into chorus), or a sharper visual direction for the custom art. Keeping changes small makes it easier to see what the model responded to and to build a shortlist of takes worth keeping.
- Genre + mood + use case in one sentence (e.g. upbeat indie pop for a product teaser).
- Lyric focus: topic, tone, and any words or names that must appear or stay out.
- Length is fixed at about 30 seconds—write hooks, not full narratives.
- Art direction in plain language if the visual will ship with the audio.
Where a 30-second AI track fits in real workflows
Lyria 3 works well for early ideation: mood boards, pitch decks, social drafts, podcast bumper experiments, and placeholder audio while a human composer finishes the full piece. Pairing lyrics with custom art is useful when you need a cohesive sample for stakeholders who will not listen without a visual frame.
It is weaker as a drop-in replacement for licensed production music or finished masters. Thirty seconds rarely covers full video timelines, live performance, or complex arrangements. Plan to export the best clip, then either loop and edit it, commission a longer arrangement, or re-prompt for alternate sections you stitch together offline.
Practical limits and good habits
Generated music and lyrics still need human review. Check that the words match your brand voice, that nothing sensitive or wrong snuck into the lyric, and that the track does not collide with existing campaigns or trademarked phrases you care about. Confirm usage rights and Gemini’s product terms for commercial use before you ship anything publicly.
Save prompts next to the clips you keep. When a 30-second sketch works, the prompt is half the asset: you can regenerate variants later or hand the brief to a composer with concrete reference audio, lyrics, and art already aligned. Used that way, Lyria 3 in Gemini is less a finished studio and more a fast drafting tool for music-plus-visual ideas.