Google DeepMind's Lyria 3 can create custom music tracks
Google DeepMind just dropped Lyria 3 in beta on the Gemini app, and it's their most advanced AI for music yet.
You can create your own 30-second tracks by typing out ideas or even uploading an image—the AI handles lyrics, music, and lets you tweak the style, vocals, and tempo for a more realistic vibe.
AI-generated audio gets a watermark
Lyria 3 is built to help you express yourself without copying real artists. It uses artist names as style inspiration rather than direct imitation, though this approach might not be foolproof.
There are filters to keep things original, plus every track gets a SynthID watermark so you know it's AI-made.
You can also check if any audio was generated using Google AI by uploading a file to Gemini, which checks for a SynthID and uses its reasoning to respond.
You can use it on YouTube too
You can use Lyria 3 in English, Hindi, Spanish, Japanese, Korean, French, German, and Portuguese. If you're a Google AI Plus/Pro/Ultra subscriber, you get higher usage limits too.
It's already available to explore on YouTube's Dream Track feature (available in the US and now rolling out to YouTube creators in other countries) to help creators add custom soundtracks to Shorts videos.