Google's Lyria 3.5 makes full vocal songs from text, images, or audio
Flow Music now edits by conversation, composes from a photo, and runs free to $64/mo — but it lives in an app, not an API.
Copy markdown
Full songs from a prompt, a picture, or a hum
Lyria 3.5 generates complete tracks — melody, lyrics, and expressive vocals — from text, an uploaded image, or recorded or uploaded audio. Point it at a mood board or a scene and get a soundtrack back.
Edit the song by talking to it
Reshape a track conversationally: shorten the intro, add a bridge, change the instrumentation or style, or adjust the mix — no re-rolling from scratch. Google says prompt adherence and structural awareness both improved this release.
What it costs, and where it runs
The free tier gives daily credit top-ups and two concurrent generations; paid plans run $8, $24, and $64 a month for roughly 600, 2,000, and 6,000 songs. It's open to adults in supported regions through a Google account at flowmusic.google.
The catch: it's an app, not an endpoint
There's no Lyria 3.5 API — it lives inside Flow Music, so you can score a demo this weekend but can't wire it into your product yet. Google's SynthID watermark travels in the audio, and the wave of AI-music licensing suits is the backdrop to watch.