Google rolled out Lyria 3 inside Gemini on February 18. Type a sentence. Get a 30-second song back, with vocals, lyrics, instruments, and cover art.

You describe what you want, it handles everything else. Google’s own framing is honest about this “The goal is not to help people make professional music. It is to give people a new way to express themselves.”

That is a small but important distinction.

How to use it explained in three steps:

1. Open Gemini. Go to gemini.google.com or open the Gemini app. Sign in to your Google account. You need to be 18 or over. If you are not, the feature will not appear.
2. Find the Music button. Look at the row of icons just below the text input box. Tap or click the one that says Create music.
3. Type your prompt and generate. Describe what you want in plain language. Hit generate. Your track and cover art appear in a few seconds. You can also upload a photo or video instead of typing. Gemini will read the mood of the image and build the music around it. Download or share directly from the app.

Tips for better results:

• Name a genre: lo-fi, afrobeat, country ballad, 90s R&B all work well.
• Describe a mood: melancholy, upbeat, nostalgic, or chaotic give the model something to work with.
• Set a vocal preference: specify male or female vocals, or ask for instrumental only.
• Reference an artist: Gemini will not copy them, but it will use their general sound as a starting point.

The limitations worth knowing

• 30 seconds only. That is the maximum track length right now. Enough for a social post or short video. Not enough for a full song.
• Still in beta. Quality will vary. Some prompts produce great results. Others will not. Adding more detail to your prompt usually helps.
• Not a professional tool. This is built for everyday expression, not music production. If you are a musician looking to replace your studio setup, this is not that.
• English-first. Eight languages are supported, but the model was primarily trained on English-language music. Results in other languages may be less polished.
• SynthID is not optional. Every track you generate will be watermarked as AI-created. That is a transparency feature, but worth knowing upfront.
• Copyright questions are not fully settled. Google says Lyria 3 was trained on music it has licensed rights to, and a deal with Universal Music Group shaped part of the approach. Filters exist to prevent copying existing artists. But the broader legal debate around AI and music is still very much active.

What is Synth ID and why does it matter?

SynthID is an invisible watermark woven into every AI-generated track at the moment it is created. You cannot hear it. It does not change how the song sounds. But a computer can detect it.

You can actually test this yourself. Upload any audio file to Gemini and ask if it was AI-generated. Gemini checks for SynthID and gives you an answer.

As AI music floods streaming platforms, having a technical way to identify what is synthetic matters more than most people realise. For artists protecting their work, for platforms managing royalties, and for listeners who just want to know what they are hearing. Google is building this infrastructure now, before it becomes a bigger problem.

Why it matters:

AI music platforms like Suno and Udio have quietly gotten very good. Full tracks that can fool most casual listeners. But they are still niche. Most people have never heard of them.

Putting Lyria 3 inside Gemini is a different move entirely. It puts AI music creation a single prompt away for the 750 million people who already use Gemini every month. People who have never thought about making music before.

That is not an incremental update. That is a distribution shift.