Gemini Now Creates Three-Minute Songs from Prompts

Google is rolling out Lyria 3 Pro inside the Gemini app, transforming quick musical sketches into full three-minute tracks. The upgraded model doesn’t just stretch the runtime; it brings a deeper grasp of composition, letting users specify structure, transitions, and instrumentation for results that sound more like finished songs than stitched-together loops.

What’s New in Lyria 3 Pro’s Music Generation Upgrade

The headline change is duration. Lyria 3 Pro increases output length from 30 seconds to three minutes, a 6x jump that finally fits standard song formats. Beyond time, the model now responds to detailed, music-native prompts. You can request a soft piano intro, a halftime chorus with stacked harmonies, or a bridge that modulates up a whole step, and the system will treat each instruction as part of a cohesive arrangement.

Google says Lyria 3 Pro understands common sections—intros, verses, choruses, and bridges—and is tuned to handle evolving themes and complex transitions. In practice, that means creators can design builds, drops, and breakdowns rather than hoping a generic vibe holds together for the full runtime.

Where You Can Use It and How Daily Limits Apply

Lyria 3 Pro is available to paying users of the Gemini app, with daily caps that scale by plan. Google AI Plus users can generate up to 10 tracks per day, Pro users up to 20, and Ultra subscribers up to 50. For developers and teams, the Lyria 3 family is also rolling out across Vertex AI, Google AI Studio, the Gemini API, Google Vids, and ProducerAI.

This makes the music model accessible whether you’re drafting soundtrack ideas on a phone or piping prompts into a production workflow for podcasts, explainer videos, training modules, or social content.

Style Boundaries, Attribution, and Training Sources

Google emphasizes that Lyria 3 Pro will not clone artists’ voices or catalog styles on command. The company says the model treats named artists as broad inspiration, not templates to mimic. According to Google, training data covers materials the company and YouTube have the right to use under terms of service, partner agreements, and applicable law.

Every Lyria output is watermarked with SynthID, Google’s content provenance signal that also marks outputs from its image and video generators. For labels, rights holders, and platforms, consistent watermarking is becoming a baseline requirement as synthetic audio circulates across streaming and short-form video.

Why Three Minutes Marks a Turning Point for AI Music

Thirty-second clips are fine for hooks, but they rarely carry a narrative arc. Three minutes makes room for the full contour of a song—intro, two verses, a chorus that returns with new energy, a bridge that reframes the theme, and an outro that lands the idea. Industry listening habits reinforce the fit: streaming-era hits often cluster around the three-minute mark, balancing replayability with enough development to feel complete.

For creators, that length also hits practical needs. Vloggers and podcasters can commission a single track that spans an entire segment, rather than looping or stitching multiple clips. Educators and marketers get room for dynamic underscores that evolve with story beats rather than repeating a four-bar motif.

How It Compares to Rival AI Music Tools and Apps

Competitors like Suno and Udio have popularized end-to-end song generation, typically capping at around two minutes with options to extend. Meta’s MusicGen and Stability AI’s Stable Audio lean more toward texture and motif creation. Lyria 3 Pro’s three-minute default, coupled with structural prompting, signals a push toward arrangement-level control rather than clip-level novelty.

The other differentiator is ecosystem reach. Being embedded in Gemini and Google’s developer stack means the same music model can serve consumer ideation, enterprise prototyping, and video tooling, which could accelerate adoption in everyday production pipelines.

Prompting Tips to Create Stronger, More Cohesive Songs

Think like a producer, not a poet. Specify tempo, key, instrumentation, and section-by-section intent. A useful example: “Uplifting synth-pop at 118 BPM in A major with palm-muted guitars and sidechain pads. Soft piano intro for 8 bars, verse with light drums, big chorus with layered vocals and clap accents, short guitar solo before final chorus, clean fade-out.”

Call out transitions where you want energy shifts—halftime feel in the bridge, filtered drums before the drop, or a key change on the last chorus. The more you describe structure and dynamics, the more Lyria 3 Pro can shape a song that evolves rather than loops.

The Bottom Line on Lyria 3 Pro’s New Three-Minute Tracks

Lyria 3 Pro moves AI music from catchy snippets to usable songs inside Gemini. With longer runtimes, structural awareness, cross-product availability, and watermarking baked in, Google is positioning its model for creators who need finished, rights-conscious tracks at scale. Whether it becomes a staple in professional sessions or a fast lane for everyday content will hinge on how convincingly it handles arrangement nuance as users push it beyond simple vibes.