How to generate songs with Gemini AI from your PC or mobile

Gemini can now generate songs

Google has taken an important step in the field of music creation with artificial intelligence. The Gemini app now includes Lyria 3Google DeepMind’s music generation model, and allows you to create tracks of up to 30 seconds from a text description or even an image. If you want to know more about the arrival of this feature, we have already told you in detail on Geeknetic. In this guide we are going to go one step further and we are going to explain exactly how to use it to generate your own song, both from your computer and from your mobile.

How to create a song with Gemini step by step

The process is simpler than it seems. The first thing is to open Gemini, either from the browser in gemini.google.com or from the application on Android. Once inside, before writing anything in the chat, you have to look at the button Toolswhich you will find in the bottom bar of the text field. Tap or click on it and, from the available options, select Create music.

This is where your creativity comes into play. You can write as free or as detailed a description as you like: the genre, the mood, the theme of the lyrics, the tempo… Gemini will accept prompts like “a melancholic pop song about a Monday morning” or something more elaborate like “an upbeat Afrobeat track with lyrics about a dog who hates the rain.”

You can also upload a photo or video so that Gemini can use them as inspiration and compose something according to what appears in the image. Of course, if you’re not feeling very inspired, you can always choose a default style.

Before sending the request, you should know that it doesn’t matter much if you use the mode Fast or Pro. While these do influence the Nano Banana, they do not influence the Lyria 3, which is a unique and specialized model. Simply send it with the mode you have selected and that’s it.

When the song is ready, Gemini presents it to you directly in the chat. Can play it instantly without leaving the application. If you like the result, you have several options to keep it: share it through a direct linkdownload it in format MP3 or download a video with cover automatically generated by Nano Banana, Google’s image model. This last option is especially practical if you want to share it on social networks as is.

All generated themes have the watermark integrated SynthIDGoogle’s system for identifying content created with AI, although it is not perceptible to the ear.

This is not a professional option

Having said all this, we must be honest: Lyria 3 on Gemini is still a very immature feature. In the tests we have carried out, it is quite difficult for him to follow the instructions precisely. If you ask for something specific, such as a specific style or a clear message in the lyrics, the result may be quite far from what you expected. It’s a fun tool to experiment with or to generate something quick to surprise someone with, but don’t expect a polished result that’s consistent with what you’ve asked for.

If what you are looking for is something closer to a real production, with control over the lyrics, style and voice, we recommend taking a look at our guide on Suno AI, which remains the reference when it comes to creating songs with artificial intelligence in a more controlled way and with much more convincing results.