Introducing Creative Space and Lyria
Short, musical meditations with any voice and any sound you describe, generated together by Google's Lyria 3.5.
We're launching Creative Space, a new part of Open Meadow for creative experiments. Its first experiment is Lyria. You choose a short meditation, describe any voice and any music you like, and Lyria generates them together as one recording.
The quickest way to understand it is to listen.
A space for experiments
Most of Open Meadow is built to be dependable. Creative Space is different. It's where we put ideas that are interesting before they're polished. Some will feel rough. Some may change a lot, or disappear.
Lyria is the first. We're also thinking about a voice designer, where you describe a voice and create it instead of choosing one from a list. Another idea is a visual map of your practice: an infographic of the meditations you've made and listened to over time.
Meet Lyria
Lyria is Google DeepMind's music model. Version 3.5 came out in early September 2026, in the Gemini app and the Gemini API. It was built to make songs, with melodies, lyrics and full arrangements. Guided meditation isn't what it was designed for, which is part of the experiment. We ask it to speak instead of sing and to leave room for silence, and it manages surprisingly well.
Our usual meditations are built in parts: we synthesize the voice passage by passage, then add pauses and music. Lyria generates the voice and the music together in a single pass, so each can respond to the other.
Short and musical
Most meditations people make in Open Meadow last around 15 minutes. Lyria recordings last up to 3. That makes them a different kind of practice: short, dense and expressive. They're more musical than our usual meditations, and the voice carries more character.
Each sample below started from a short meditation script. What changes is the voice and sound we asked for.
Any voice, any music
When you pick a meditation, Open Meadow suggests a voice and a background sound that suit what your script is about. A grounding practice might get a low, steady voice over a frame drum. A morning practice might get a bright voice and birdsong.
Then the prompt is yours to change. Ask for an accent, an age, a texture or a persona. Ask for an instrument, a genre, a place, or no music at all and only rain. You can edit the script too.
Expect a few glitches
Lyria is new, and we're asking it to do something it wasn't designed for. It shows. Here's what we've noticed so far:
- Now and then, the narrator drifts into singing.
- It sometimes skips or changes a few words of the script.
- It doesn't always follow a direction, like a pause or a specific instrument.
- It occasionally gets the voice wrong, for example a man's voice when you asked for a woman's.
- Pauses and timing vary from one recording to the next.
If a recording doesn't come out the way you hoped, generating another version often gives a different result. And sometimes the unexpected version is the one you keep.
How it works
- Up to 3 minutes. Lyria works with meditations of about 3 minutes or less. If you don't have one, ask Omi in Chat for a short meditation, then open Lyria.
- Made from your meditations. Choose one from your library. Your saved meditation doesn't change.
- Listen and download. Your recordings stay on the Lyria page. Play them any time, or download them as MP3 files.
- 1.5 credits per recording. Lyria is available with a subscription, during your trial, or with extra credits. See Plans and Credits for details.
Come and experiment
We made Creative Space so we can explore together. Try a voice you would never pick from a list, or a sound you didn't know you wanted. Then tell us what you made, and what you'd like to see here next.