Gemini opens Lyria 3 music
Google gave full access to Lyria 3 music generation inside Gemini, letting users generate up to five full-length tracks per day for free. That move signals major cloud vendors are treating creative, high-bandwidth multimodal features as first-class consumer offerings inside broader AI assistants. (x.com)
Google just moved music generation from a demo page into the same chat box people already use for writing, search, and images. In Gemini, users can now make songs by typing a prompt or uploading a photo, and Google says the feature began rolling out in beta on February 18, 2026. (blog.google) At launch, Gemini’s music tool made 30-second tracks, which is short enough for a joke song, a podcast intro, or a social post. Google’s help page says anyone 18 or older with a personal Google account can use that basic version, while longer tracks require a paid Google artificial intelligence plan. (support.google.com) Then Google widened the gate. On March 25, 2026, it announced Lyria 3 Pro, which can generate tracks up to 3 minutes long and lets users ask for specific parts like an intro, verse, chorus, or bridge. (blog.google) That changes the product from “make me a quick jingle” to “draft me an actual song.” Google says Lyria 3 Pro is available not just in Gemini, but also in Vertex artificial intelligence, Google artificial intelligence Studio, the Gemini application programming interface, Google Vids, and ProducerAI. (blog.google) The free tier is what makes this feel different from a lab launch. Google’s support pages say Gemini users can create music directly in the app, download the result as an MP3 audio file or MP4 video file, and share it with a link, while subscription tiers raise the limits and unlock longer tracks. (support.google.com) Google is also treating the songs like content that will spread far beyond Gemini itself. The company says every generated track includes SynthID, its watermarking system for identifying Google-made audio, and the Gemini help page says the watermark is embedded in each track. (blog.google, support.google.com) Under the hood, Lyria 3 is not a toy soundboard that stitches loops together. Google DeepMind’s model card says it is a music generation system that turns text into audio and lyrics, and that it is distributed through Gemini, Vertex artificial intelligence, Google artificial intelligence Studio, Google Vids, YouTube Producer AI, and Music AI Sandbox. (deepmind.google) That distribution list is the real tell. Google is putting the same music model into consumer chat, developer tools, cloud products, and video software at once, which is the software equivalent of putting the same engine in a family sedan, a work truck, and a race car. (deepmind.google, blog.google) The old pattern in artificial intelligence was to keep creative tools in separate apps with separate brands. Google is doing the opposite: it is folding music into Gemini, the same assistant it already uses for text, images, and task help, so “make me a song” sits beside “summarize this document.” (blog.google, blog.google) That is why this launch is bigger than one music model. When a company gives everyday users free song generation inside its main assistant, and gives developers the longer-track version through the same stack, it is betting that high-bandwidth media creation will be a standard feature of general-purpose artificial intelligence, not a niche add-on. (blog.google, deepmind.google)
