Google's Lyria 3.5 music model is on the Gemini API at $0.08 per song

Google made Lyria 3.5 (lyria-3.5), its music generation model, generally available on the Gemini API on September 3, 2026. It writes full-length songs with vocals, timed lyrics and instrumental arrangements from a text prompt, and can take images as well. Google’s pricing page lists it at $0.08 per song, with no free tier.

What it makes

• Length: the docs give “a couple of minutes”, set through the prompt (for example, “create a 2-minute song”). They give no hard maximum.
• Lyrics: generated in the language of your prompt, or you supply your own with section tags such as [Verse] and [Chorus].
• Inputs: text, plus up to 10 images with a prompt.
• Output: 44.1 kHz stereo, as MP3 by default or WAV on request.
• Watermark: every track carries a SynthID audio watermark, which Google says can’t be heard.

For short pieces, the docs also list Lyria 3 Clip (lyria-3-clip-preview). It always makes a 30-second MP3 clip and is meant for loops, previews and short clips.

What it costs

Model	Price	Free tier
Lyria 3.5	$0.08 per song	No

The price is per song, not per second, on Google’s pricing page as we read it on October 18, 2026.

Who it’s for

Lyria 3.5 is an API model for developers who want to build music into an app or a video pipeline. It isn’t a consumer app with plans like Suno. You pay per song through a Gemini API account.

The pages we read don’t cover licensing or commercial-use terms for the songs. Read the Gemini API terms before you use generated music in paid work, and keep in mind that every track carries a SynthID watermark.