Google Adds Lyria 3 AI-Music Model to Its Gemini App

Why It Matters

The move positions Google as a major player in AI‑driven music production, opening new revenue streams while addressing industry concerns over copyright and attribution.

Key Takeaways

• Gemini reaches 750 M monthly active users.
• Lyria 3 creates 30‑second AI‑generated tracks.
• Prompt with text, photo, or video for music.
• Google adds copyright filters and SynthID watermark.
• Dream Track expands AI soundtracks for YouTube Shorts.

Pulse Analysis

Google’s decision to embed Lyria 3 within Gemini marks a strategic shift toward democratizing AI‑generated music. By leveraging Gemini’s massive user base, the company transforms casual content creation into a seamless audio‑first experience. Users can describe a mood, upload visual media, and receive a polished, 30‑second track in seconds, blurring the line between professional composition tools and everyday apps. This integration also showcases Google’s broader vision of a unified generative AI ecosystem, where text, image, video, and now sound converge under a single interface, enhancing cross‑modal creativity for both consumers and creators.

The technical rollout balances innovation with practical constraints. Lyria 3’s output is limited to short, royalty‑free clips, complemented by Nano Banana’s AI‑generated cover art, ensuring quick turnaround without overwhelming storage or processing resources. The model’s expansion into YouTube’s Dream Track tool extends its reach to short‑form video creators, offering automated soundtracks that adapt to visual cues. Google’s safeguards—including style‑inspired prompts rather than direct imitation, content‑matching filters, and the SynthID watermark—aim to mitigate infringement risks while maintaining user flexibility. These measures reflect a growing industry consensus that responsible AI deployment must be baked into product design.

For the music and tech sectors, Lyria 3’s Gemini integration signals intensified competition in the AI‑music market, traditionally dominated by startups and niche platforms. Google’s scale and licensing framework could pressure rivals to adopt similar compliance standards, potentially reshaping royalty structures and creator compensation models. Moreover, the ease of generating short, brand‑safe tracks may accelerate adoption of AI music in advertising, gaming, and social media, driving new monetization pathways. As AI-generated audio becomes mainstream, stakeholders—from record labels to independent artists—will need to navigate a landscape where creativity, technology, and legal frameworks intersect more tightly than ever.

Google adds Lyria 3 AI-music model to its Gemini app

By Stuart Dredge · February 18, 2026

Gemini has become the flagship service for Google’s GenAI tools: a multimodal AI assistant that can answer questions, conduct research, automate tasks and create images and video clips among other tasks.

Now it can generate music too. This afternoon, Google has added the latest version of its DeepMind team’s Lyria generative‑music model to Gemini, thus putting it at the fingertips of many more users.

How many? In parent company Alphabet’s recent earnings call, CEO Sundar Pichai said that the Gemini app had more than 750 million monthly active users. A blog post today announced how those people will be able to use Lyria 3 to make music.

“Just describe an idea or upload a photo, like ‘a comical R&B slow jam about a sock finding their match’ and in a matter of seconds, Gemini will translate it into a high‑quality, catchy track,” the post explained.

Users can also upload a photo or video and ask Gemini to generate a track with lyrics that match its mood. Note, there are limits. For now, Gemini can only make 30‑second tracks with Lyria 3, tapping Google’s Nano Banana AI to generate cover art.

“The goal of these tracks isn’t to create a musical masterpiece, but rather to give you a fun, unique way to express yourself,” the blog post added.

Alongside the Gemini news, Google also announced that Lyria 3’s integration with YouTube’s Dream Track tool is expanding beyond the US. It enables YouTubers to generate AI soundtracks for their Shorts videos.

As with any AI‑music announcement from a giant technology firm, we have some questions about training, as well as respect for artists and copyright.

The blog post presented Lyria as something where “we’ve sought to develop this technology responsibly in collaboration with the music community”, adding that Google has “been very mindful of copyright and partner agreements as we’ve trained Lyria 3”.

What does that mean in practice? Music Ally understands that the policy being applied here is that Lyria’s training is designed to use music that YouTube and Google have the rights to use under their terms of service, partner agreements and “applicable law”. That’s a phrase for lawyers to discuss…

Back in early 2024, Billboard reported that when Google DeepMind was first working on music, it “trained its model on a large set of music — including copyrighted major‑label recordings — and then went to show it to rights holders, rather than asking permission first”.

However, YouTube has partnered with right‑sholders and artists on various GenAI experiments, winning praise in 2023 from Warner Music Group CEO Robert Kyncl for its approach – obligatory context: he spent 12 years at YouTube before moving to WMG.

Lyria 3’s Gemini integration has some additional guardrails that will be welcomed by music right‑sholders, as outlined in today’s announcement.

“Music generation with Lyria 3 is designed for original expression, not for mimicking existing artists. If your prompt names a specific artist, Gemini will take this as broad creative inspiration and create a track that shares a similar style or mood,” it explained.

(That does raise questions about what the model has been trained on in order to understand what styles or moods are similar to specific artists. However, this is information that could be gleaned from a variety of sources, not just the original music.)

“We also have filters in place to check outputs against existing content,” added Google, while flagging the ability for people to report “content that may violate your rights or the rights of others”.

Finally, all tracks generated within Gemini will have Google’s SynthID watermarking embedded in their files. There is also a feature that will enable Gemini to recognise those tracks, just as it can already identify images and videos made using Google’s GenAI tools.

“Simply upload a file and ask if it was generated using Google AI, and Gemini will check for SynthID and use its own reasoning to return a response,” the blog post explained.