Google Puts Lyria 3.5 AI Music Maker in Gemini

Google launched Lyria 3.5 AI music in Gemini and its API worldwide, dodging the copyright suits plaguing Suno and Udio.

Google has rolled out Lyria 3.5, which it calls its best-sounding music generation model, into the Gemini app, the Gemini API, Google AI Studio and Google Vids. Users pick or describe a genre, choose sung or instrumental output, and select templates for jobs like video backing tracks, brand jingles, ringtones and birthday songs. Tracks run up to about three minutes in high-fidelity stereo.

According to We Rave You, the surprise is what is not attached to the launch: Suno and Udio are both defending copyright suits from labels, artists and collecting societies, while Google enters the same market without that legal weight. Pasquale Pillitteri reports the timing lands just after a Munich Regional Court ruling on July 31, 2026 found Suno liable for copyright infringement, holding that protected works were stored and reproducible in its models and rejecting the argument that users bore responsibility through their prompts. Google's new guardrails — refusing to mimic specific artists' voices and blocking copyrighted lyrics — match precisely the points on which Munich convicted Suno, the outlet notes, though Google has not said the two are connected.

Google describes improvements across musicality, singing, lyrics and length: more natural melodies, better pronunciation and emotional nuance, closer prompt adherence, and full songs with verses, choruses and bridges rather than loops. Pasquale Pillitteri reports the API exposes two variants — lyria-3.5 for full tracks with prompt-controlled duration, and a clip variant producing fixed 30-second snippets — outputting 44.1kHz stereo audio in MP3 or WAV, with support for up to 10 images as multimodal input and tagged lyrics with timestamps. Every track carries a SynthID watermark identifying it as AI-generated or modified.

We Rave You argues a well-resourced third entrant changes the shape of a market previously defined by two companies fighting on multiple legal fronts, because Google can absorb legal risk that a startup cannot. The same outlet reports AI already generates more new music than humans, and Google entering at this scale accelerates that. Pasquale Pillitteri cites Deezer figures that more than half of new uploads are now AI-generated — roughly 90,000 fully synthetic tracks a day — with real listenership stuck at 1–3%, making the watermark a distribution-side filter as much as an ethical signal.

We Rave You reports Google has not detailed what Lyria 3.5 was trained on in the material accompanying the Gemini release, noting that a cleaner legal position today is not the same as a disclosed one. The outlet observes that engineers debating whether AI replaces producers keep returning to the same point: what matters is not the quality of the output but whether anyone can say where it came from. Neither We Rave You nor Tbreak UAE reports any separate pricing announcement for music generation, with API use falling under normal Gemini API billing according to Pasquale Pillitteri.

For producers, the tools are no longer a two-horse race, and We Rave You suggests the terms on which AI music gets made will now be negotiated by a company with considerably more leverage. The Munich ruling against Suno can be appealed and is not final, according to Pasquale Pillitteri, so the legal contest that shaped the market so far remains open. Meanwhile Google positions Lyria 3.5 for quick creative jobs — backing tracks, jingles, ringtones — rather than replacing a recording studio, with Tbreak UAE noting the prompt box remains the main instrument.