Top AI Product

So Google dropped [Lyria 3](https://blog.google/innovation-and-ai/products/gemini-app/lyria-3/) yesterday, and honestly, I wasn’t expecting to spend my evening typing silly prompts into Gemini and bobbing my head to AI-generated tracks. But here we are.

Lyria 3 is Google DeepMind’s latest music generation model, and it’s now baked right into the Gemini app. The pitch is simple: type what you want to hear, and it spits out a 30-second track complete with vocals and lyrics. You can tweak the style, tempo, and vocal feel, and unlike the older Lyria models, you don’t need to write your own lyrics anymore — it handles that for you based on the prompt. Ask for “a comical R&B slow jam about a sock finding their match” and yes, it will deliver exactly that.

But the thing that really caught my attention is the multimodal angle. You can upload a photo — say, a sunset from your last trip — and Lyria 3 will compose something that fits the mood. Throw in a birthday party video and it’ll generate a soundtrack for it. That’s a genuinely fun party trick, and I could see people actually using this for Instagram stories or quick YouTube Shorts. Speaking of which, Lyria 3 is also [rolling out to Dream Track on YouTube Shorts](https://deepmind.google/models/lyria/) for creators who want custom background music.

The model supports eight languages including English, Spanish, French, Japanese, Korean, and Hindi, so this isn’t just a US-centric launch. It’s going out to all Gemini users 18 and older worldwide, with higher usage limits for AI Plus, Pro, and Ultra subscribers.

On the responsibility side, every track gets stamped with [SynthID](https://deepmind.google/models/lyria/), Google’s invisible watermark for AI-generated content. You can even upload a track to Gemini and ask if it was made by AI — it’ll check for the watermark and give you an answer. That’s a nice touch given all the discourse around AI-generated media lately. Google also made it clear that the model is designed for original expression, not for cloning existing artists, which feels like a deliberate move to sidestep the copyright headaches that have plagued other music AI tools.

The [Hacker News threads](https://news.ycombinator.com/item?id=47062463) are already buzzing, and coverage from [TechCrunch](https://techcrunch.com/2026/02/18/google-adds-music-generation-capabilities-to-the-gemini-app/) and [The Verge](https://www.androidcentral.com/apps-software/geminis-music-generator-is-here-and-i-think-this-is-where-everyday-ai-gets-interesting) has been pretty extensive. It’s the first time a major chat platform has shipped a consumer-facing music generator this directly — no separate app, no waitlist, just ask Gemini to make you a song. Whether the novelty sticks or fades after a week of joke tracks, having this built right into a tool millions of people already use is a pretty big deal.