Ad
Skip to content

Google brings AI music generation to Gemini with Deepmind's Lyria 3

Image description
Google Deepmind

Key Points

  • Google is integrating Lyria 3, an AI music generator developed by Google Deepmind, directly into the Gemini app, allowing users to create music from simple text prompts.
  • The model generates 30-second tracks complete with vocals, lyrics, and cover images based on text descriptions or uploaded photos and videos.
  • The feature is now available on desktop and will soon roll out to the mobile app for all users aged 18 and over, supporting eight languages.

Google Deepmind's new Lyria 3 model creates 30-second tracks with vocals, lyrics, and cover art from a simple text prompt or uploaded media.

Users describe a genre, a mood, or a personal memory, and Gemini spits out a finished 48kHz stereo track. Compared to earlier Lyria models, Lyria 3 writes lyrics on its own, offers more control over style, voice, and tempo, and produces more musically complex output. It also works with uploaded photos and videos as input.

The feature is live today on desktop and hits the mobile app over the coming days, according to Google. It's open to all users 18 and over in eight languages, including German and English. Google AI Plus, Pro, and Ultra subscribers get higher usage limits. Google hasn't specified exactly which countries have access through the Gemini app.

When a user names a specific artist in their prompt, Gemini is supposed to treat their style as loose inspiration rather than copy it. Google says "Music generation with Lyria 3 is designed for original expression, not for mimicking existing artists," filters check generated tracks against existing content, and users can flag potential infringements. Every generated track gets a SynthID watermark to tag it as AI-generated.

Ad
DEC_D_Incontent-1

AI music is getting too real to stay "just for fun"

Google is careful to address copyright concerns upfront, and there's good reason for that caution. The rapid progress in AI music generation and early streaming chart hits, along with lawsuits from Sony Music and other labels, show that AI music is already far more than a novelty. Google frames the feature as "for fun" and says it was "mindful" of copyrights and partner agreements when training Lyria 3. What that means in practice is unclear, since the company doesn't disclose specific training data or commercial deals.

OpenAI is reportedly building a music generator for ChatGPT. For now, Suno remains the only widely available high-quality option. Its competitor, Udio, was acquired by Universal late last year and is still available, but doesn't seem to be improving its platform anymore.

AI News Without the Hype – Curated by Humans

As a THE DECODER subscriber, you get ad-free reading, our weekly AI newsletter, the exclusive "AI Radar" Frontier Report 6× per year, access to comments, and our complete archive.