Lyria Cheat Sheet
Last updated: April 2026
Quick Facts
Pricing
Free for now via experimental platforms like MusicFX and AI Test Kitchen. No official pricing for a standalone product has been announced.
Free Plan
Yes + includes full access to text-to-music and melody-to-music generation within the experimental interfaces, with download options.
Rating
4.1/5
Best For
Musicians and creators who want to quickly generate high-fidelity, structured musical ideas from text or humming, not for full production.
Key Features
- ✓Text-to-Music Generation
I tested this extensively. You type a descriptive prompt like 'upbeat synthwave with a melancholic melody' and get a surprisingly coherent 30-60 second instrumental track.
- ✓Melody-to-Music (Hum to Search)
In my experience, this is a game-changer. Hum or whistle a tune into your mic, and Lyria builds a full instrumental arrangement around it, perfect for developing ideas.
- ✓High-Fidelity Audio Output
What surprised me was the audio quality. It's not MIDI; it's rendered audio with realistic instrument textures and spatial depth, rivaling basic sample libraries.
- ✓Structured Compositions
Lyria doesn't just loop. It often creates clear intro, verse, and chorus sections with dynamic variation, which most AI music tools struggle with.
- ✓Expressive Vocal Generation
Through Dream Track, I heard AI-generated vocals in the style of specific artists. The expressiveness and timbre are uncanny, though public access is limited.
- ✓Genre Versatility
I've successfully prompted everything from classical string quartets and lo-fi hip-hop to aggressive drum and bass. Its genre comprehension is impressively broad.
- ✓Instrument Control
You can specify instruments in your prompt (e.g., 'acoustic guitar, piano, cello'). Lyria handles these requests well, creating believable mixes between them.
- ✓Mood & Emotion Prompting
Using words like 'joyful,' 'somber,' or 'tense' actually influences the harmonic and melodic output. It's more than just a keyword filter.
- ✓Fast Iteration
Generations take 20-40 seconds. This speed lets you rapidly iterate on a prompt, which is crucial for finding a usable musical idea.
- ✓Integration with YouTube Shorts
The Dream Track experiment let creators generate short music clips in an artist's style for Shorts. It was a slick, if gated, implementation.
- ✓Google AI Test Kitchen Access
This is the main playground. The interface is simple: a prompt box, a generate button, and a library of your past creations for download.
- ✓Developer API Potential
As a model from DeepMind, the real power is under the hood. I expect a future API for developers to build custom music apps on top of Lyria.
Tips & Tricks
Be hyper-descriptive. Instead of 'sad song,' try 'a slow, minimalist piano piece in C minor, with a weeping cello line entering in the second half.'
Use the hum-to-music feature to break out of generic results. Your own simple melody gives Lyria a unique seed that produces far more original output.
Specify tempo and key when you can. Prompts like '120 BPM funk track in F major' yield more targeted and usable results for music production.
Chain prompts. Generate a basic track, hum a part of it back in, and let Lyria reinterpret it to create variations for a longer arrangement.
Download everything you like. Access is experimental and could change or be put behind a paywall. Build a library of stems and ideas now.
Limitations
- -You cannot edit the generated audio stems individually; you get a single mixed audio file, limiting production flexibility.
- -Output is capped at about 70 seconds, making it unsuitable for generating full-length song structures without manual editing.
- -Access is fragmented and indirect through experiments like MusicFX, not a dedicated, full-featured Lyria application or DAW plugin.
- -While vocals are impressive, public tools for generating custom vocals (beyond artist experiments) are not readily available.
- -There is no fine-grained control over mix levels, song structure, or chord progressions—you are at the mercy of the prompt.