Describing a genre in a sentence is now enough to get a finished song inside Google’s Gemini app. Lyria 3.5, the company’s music model, moved into the assistant and its API on September 4, after debuting earlier in the summer inside the Flow Music tool. Google calls this iteration its best-sounding music model, pointing to more expressive vocals, denser arrangements, and higher output fidelity.
Gemini users get a guided creation flow. They can type or select a genre, choose vocal or purely instrumental output, and build from templates aimed at backing tracks for video, brand jingles, or personalized ringtones, with track length adjustable from short clips to full songs. The same model is also reachable through Google AI Studio and Google Vids, while Flow Music keeps serving artists.
The underlying generator produces 44.1 kHz stereo audio from text or even an uploaded image, and it keeps vocals, timed lyrics, and instrumentation coherent across a piece that can run to about three minutes. Developer docs split the family into two endpoints: lyria-3-clip-preview, which always returns 30-second loops, and lyria-3.5, which writes complete songs with verses, choruses, and bridges.
The placement inside Gemini is the strategic part. Google now sells music generation through the same app where people chat with its flagship assistant, priced through standard Gemini access rather than a separate music subscription. That puts a mainstream audience one prompt away from custom composition, a direct challenge to the standalone AI music services that spent the past year building their own followings.
