You hit record on a late-night clip, half-expecting a blooper. Gemini returned a 30-second soundtrack that sounded uncannily right and entirely automatic. I felt a thrill — and the same instant, a question: what just happened to the middleman?
I’ve spent days testing Gemini’s new music tools so you don’t have to. I’ll walk you through what works, what feels like marketing, and how this changes the way you make short music for video, social, or a quick idea sketch.
At my kitchen table I dropped in an old vacation clip and it spit back a soundtrack — what Lyria 3 actually does
Google’s Lyria 3 is now powering audio generation inside the Gemini app. You can feed it a text prompt, a photo, or a video and Gemini returns a 30‑second composition: instrumental beds, AI-sung parts, and even lyrics. I tried mood prompts, told it to sound melancholy and acoustic, then swapped the vocal color to “breathy R&B” — the results shifted to match.
Google calls Lyria 3 its most advanced music-generation model so far. It composes across genres — 90s rap, Latin pop, Folk ballad, R&B romance — and it will shape the vocal timbre and lyrical tone if you ask. For quick idea work, it feels like a Swiss army knife for song ideas.
How do I make music with Google Gemini?
Open the Gemini app, tap Tools, then choose Create music. Give the model a prompt (text, image, or video), optionally paste lyrics or sing a short sample, pick a genre, and press generate. In my tests each render stays inside the 30‑second limit and comes back fast enough to keep you iterating.
On creators’ feeds I saw Dream Track tags already — distribution and where you’ll find the feature
Creators I follow began tagging short clips with Dream Track within hours of the rollout. Gemini’s music tool is rolling out broadly and free inside the app; Google has also integrated Lyria 3 into YouTube’s Dream Track so creators can drop AI-generated audio straight into Shorts and videos.
To reach it in Gemini, go to Tools > Create music. You can remix an existing clip or let Gemini produce a fresh 30‑second piece. The tight length makes it ideal for social clips and hooks, not full songs — think of each output as a sketch rather than a finished album cut.
Is Gemini’s music generation free?
Yes. Google is deploying the feature for free inside Gemini. The trade-off is the 30‑second cap per generated track, which keeps it friendly for rapid prototyping but limited for long-form production.
A label exec flinched when I played an AI vocal — the safeguards and the competition
A friend who runs licensing heard an AI vocal and immediately asked whether it copied anyone he knew. Google says Lyria 3 includes filters to prevent outputs that mimic identifiable artists, and the company frames the tool as a creative assistant rather than a replacement for musicians.
That positioning matters. Suno currently leads among consumer-facing AI music tools by offering longer generative tracks — typically two to four minutes — while reports place OpenAI somewhere down the road with music ambitions of its own. Lyria 3’s short-form focus slots it into idea generation and creator tools rather than album production, and the guardrails make it safer for platforms and publishers.
In practice the model is useful when you need a quick hook or an atmospheric bed for a short video. It’s a paintbrush dipped in radio waves — effective for strokes, not whole murals.
Can Gemini generate vocals and lyrics?
Yes. Gemini can produce instrumental tracks and add AI-generated vocals with lyrics you supply or that it writes for you. You can tweak the vocal style and the mood, then re-run the generation until it fits the clip or the hook you’re chasing.
This is the moment where you choose: will you seed your edits with 30‑second AI sketches in Gemini, keep using longer generators like Suno for fuller tracks, or wait to see what OpenAI does next?