Google launched Lyria 3 on February 18, 2026 — its most advanced AI music generation model to date — directly inside the Gemini app for users worldwide. Built by Google DeepMind, Lyria 3 marks a decisive shift from cautious experimentation to confident deployment: after years of limited-access music AI tools like MusicLM and Lyria 2, Google has now bundled professional-grade music generation into the same app that hundreds of millions of people use for chat, image generation, and research. The launch makes Google the first major tech company to integrate music generation natively into a general-purpose AI assistant, putting it in direct competition with dedicated music AI startups like Suno and Udio at a scale neither could match overnight.
Lyria 3 is developed by Google DeepMind, the same research division behind Gemini's core reasoning improvements and the Veo video generation models. The model uses a combination of deep learning architectures including transformers and recurrent neural networks to learn intricate patterns in music — not just chord progressions and rhythms, but the relationships between mood, instrumentation, tempo, vocal style, and genre that make a piece of music feel cohesive. DeepMind's internal team describes Lyria 3 as a "new instrument" designed for collaborative co-creation, not a replacement for musicians, but a tool that lowers the barrier to musical expression for people who have ideas but lack technical training. The goal, as stated by Joël Yawili and Myriam Hamed Torres in Google's official blog post, is not to create a musical masterpiece but to give users "a fun, unique way to express yourself."
Lyria 3 improves on its predecessors in three concrete ways. First, users no longer need to supply their own lyrics. The model generates lyrics automatically based on the user's text prompt, inferring the appropriate words, tone, and phrasing from the mood and genre described. This removes one of the biggest practical barriers to music creation: most people can describe what they want to feel, but far fewer can write lyrics that match. Second, users now have more granular creative control over musical elements. The interface lets you steer style, vocal character, tempo, instrumentation, and acoustic preferences — meaning you can push toward a specific sound rather than accepting whatever the model defaults to. Third, Lyria 3 produces more realistic and musically complex tracks than earlier versions, with tighter alignment between the described mood and the actual audio output, better vocal believability, and more sophisticated arrangements that hold together as real compositions rather than sound collages.
From a practical standpoint, using Lyria 3 inside the Gemini app is simple. You describe a genre, mood, scenario, or vibe in natural language — anything from "a lo-fi jazz beat for studying" to "an epic orchestral theme for a fantasy battle" or "a birthday jingle for my dog" — and Lyria 3 generates a fully produced 30-second track complete with vocals and lyrics. You can also upload a photo or a short video clip, and the model will compose an original soundtrack that matches the mood and tone of whatever you shared. Each generated track comes with automatically created cover art powered by Google's Nano Banana image generation model, and every output is watermarked with SynthID — Google DeepMind's imperceptible AI-content signature embedded directly into the audio — so the track is permanently identifiable as AI-generated content without affecting the listening experience.
Access tiers are structured around Gemini subscription plans. Free Gemini users can generate music tracks, but with lower usage limits. Google AI Plus, Pro, and Ultra subscribers receive higher generation quotas, enabling more tracks per session and priority access during peak demand. The rollout began on February 18, 2026 as a full global release across Rapid Release and Scheduled Release domains, with complete availability across all regions expected within one to three days. The feature is available in English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese at launch, with Google planning additional language expansions as the model matures. Users must be 18 years of age or older to access the music generation feature.
Beyond the consumer Gemini app, Lyria 3 is simultaneously being integrated into YouTube's Dream Track feature for Shorts creators in the United States, with international expansion underway. YouTube Shorts creators can use Dream Track to generate original music for their videos directly within YouTube Studio, removing the need to license third-party music or record original audio. This Shorts integration extends the reach of Lyria 3 far beyond the Gemini app itself, effectively putting DeepMind's music model into the hands of YouTube's creator economy — one of the largest content production ecosystems on the planet. A separate YouTube tool called Music Assistant, part of the Creator Music marketplace, lets YouTube Partner Program members generate copyright-free instrumental backing tracks via text prompts, and the Lyria 3 upgrade enhances that capability as well.
Copyright and training transparency are sensitive areas for any AI music model, and Google has addressed this directly. DeepMind states that Lyria 3 was designed for original expression rather than artist imitation, and that prompts naming specific artists will be treated as broad stylistic inspiration rather than direct mimicry. Output filters check generations against existing content, and Google states that Lyria 3 was trained on music that YouTube and Google "have the right to use" under their terms of service, partner agreements, and applicable law. The company acknowledges it has been "very mindful of copyright and partner agreements" throughout development and says it has worked to "develop this technology responsibly in collaboration with the music community." Every track generated through Gemini carries a SynthID watermark, and Google has extended SynthID verification to audio uploads — meaning users can check whether any given audio file was produced by a Google AI model.
In competitive context, Lyria 3's Gemini integration arrives as AI music generation enters a pivotal phase. Suno and Udio established the category with consumer-friendly tools, but both remain standalone apps with no connection to broader AI ecosystems. Google's integration of Lyria 3 into Gemini leverages a platform advantage neither startup can replicate: hundreds of millions of existing Gemini users now have music generation one prompt away, alongside text, image, video, code, and research capabilities. Anthropic's Claude and OpenAI's ChatGPT have no comparable music generation built in, making this a genuine multimodal differentiator for Google's AI assistant strategy. For creators, marketers, and everyday users, the practical implication is that generating an original jingle, a birthday song, a podcast intro, or a custom soundtrack for a video no longer requires specialized tools, subscriptions, or creative skills — it requires only a description and a few seconds of waiting.