Lyria 3 Pro

What it does

Lyria 3 Pro is Google's text-to-music model, served on pipe2 through the Music Generator pipeline. You give it a written description of a mood, genre, and length; it returns an original audio track. There is no audio, image, or MIDI input: text in, music out. Its stated character is ambient pads and slow textures, which is why it is pointed at documentary-style scoring.

When to use it

• Running Music Generator to score a documentary, explainer, or interview cut where the bed needs to sit under speech instead of competing with it.
• You need original, royalty-free background music and want to skip stock-library licensing entirely.
• The brief is atmospheric: ambient pads and slow textures. That is the material this model is described for.
• You can describe what you want in words: mood, genre, and length are the inputs Music Generator takes.
• You want a track back in seconds rather than commissioning or hand-editing one.

What to know before you run it

This model is billed on pipe2 in two shapes, and the live price table on this page shows the current rates for each. There is a fixed 30-second clip unit, and a flat per-song unit that covers a full generated track. The clip tier is the shorter, fixed-length commitment; the per-song tier is the one to pick when you need a usable full-length bed. Because per-song pricing is flat, a longer song does not scale the charge the way a per-second model would, but you also cannot trim the cost by asking for slightly less.

No reference image or seed audio is accepted, so the prompt is your only control surface. There is one published example on this page; play it to hear the model's default texture before you spend on a longer generation.