Describe a mood, genre, or story in plain words, and Lyria AI turns it into a complete song. No musical training needed; get a full track, ready to use.
Try Lyria AI For Free
Making music with Lyria AI works differently from typing a prompt into a basic generator. It reads your words closely and builds them into a full song with structure and vocals.
Lyria 3.5 creates cohesive songs with natural flow from start to finish, with tracks lasting up to 3 minutes. It supports vocals, lyrics, and detailed prompts, giving each song more structure and creative control for you.
Lyria RealTime creates a continuous stream of music that responds as you change its controls. Adjust tempo, instruments, mood, key, density, or brightness while the music keeps playing without stopping or restarting as needed.
Upload a photo instead of a text prompt. Lyria reads the mood and colors in it and creates a matching track. A sunset might turn calm and ambient while a busy street scene comes out upbeat and energetic.
Every track the tool makes carries a SynthID watermark inside the audio itself, so it can always be identified as AI-made. The watermark stays inaudible and does not change how the track sounds to listeners.
Musicians want a track that actually sounds finished, and developers want a model they can create on. Here's why both keep coming back to Lyria AI:
With Lyria, music becomes something you can shape as it plays, rather than a one-time request. Listen, adjust, and guide the track in real time, giving you more creative control than simply typing a prompt.
Try Lyria AI For Free
Access Lyria across Google tools, including Vertex AI, Google AI Studio, the Gemini API, the Gemini app, and Google Vids. Developers can also integrate supported music capabilities into their own apps and creative projects.
Try Lyria AI For Free
Lyria includes safety measures for generated music, along with SynthID watermarking and content labels. These tools help identify AI-made audio and support safer use, while keeping the generated tracks useful for creative work.
Try Lyria AI For Free
Lyria AI isn't the only music model available, so it's worth seeing where it stands next to the rest. Here's how it compares with other music tools on the features that matter most:
| Platform | Longest Output | Vocals | Real-Time Control | Users |
|---|---|---|---|---|
| Lyria AI | 3 minutes of audio | Timed lyrics included | Yes, through RealTime | Developers and studios |
| Magiclight AI | 50 minutes of video | 37 voices, 11 languages | No, storyboard editing | Story creators |
| Udio | Full songs | Yes, with lyrics | No | Songwriters |
| Suno | Full songs | Yes, with lyrics | No | Hobbyists |
| Soundful | Background beds | Instrumental only | No | Content creators |
Creating a track with Lyria AI is simple once you know the steps. Learn how to go from an idea to a finished song through this guide.
Choose Lyria 3 Pro for a structured song, Lyria 3 for a 30-second asset, or RealTime for anything interactive.
Go to the Gemini app or native Google AI Studio. Describe the mood and genre, or upload a reference image and let the model read the style from it. Add your own lyrics here if the track needs specific words.
Set the duration and section structure on a composed track. On a live stream, pull the sliders for key, tempo, density, and brightness as it plays.
Hit generate to create your track depending on the specific Lyria version used. Preview it and download it as an MP3 or video file.
Game Audio Creator
Adaptive scores used to mean crossfading between prerecorded stems. Lyria RealTime shifts the music with the gameplay instead, and players stop noticing loop points.
App Developer
Forking a demo in AI Studio got me a working prototype before lunch. Wiring Lyria into the app took less time than picking a licensing library would have.
Music Producer
I use Lyria for ideas rather than finished tracks. Changing the key while a stream is running shows me where a piece could go faster than sketching it out myself.
Video Content Creator
Every clip needed a music bed that would not get claimed. Lyria AI writes one to the brief, and the watermark means nobody argues about where it came from.
Google DeepMind's family of music generation models, not a standalone app. Access it through Vertex AI, Google AI Studio, the Gemini API, the Gemini app, and Google Vids rather than signing up on a separate website.
Lyria 3 Pro composes up to 3 minutes with proper structure, while the standard Lyria 3 stops at 30 seconds. RealTime is different again, since the stream simply continues until you close it.
Yes, vocals and timed lyrics come as standard, and you can supply your own words for the model to sing instead. Add “instrumental” to the prompt when you only need a bed under a video.
It never stops; rather than planning a whole song and rendering it, the model streams continuously and responds while playing. Shift genre, mute instruments, or change tempo without a break.
Yes, with SynthID, embedded inaudibly so the listening experience is unaffected. That watermark lets anyone confirm later that a piece of audio came from an AI model rather than a session player.
Lyria AI has no consumer app, so casual users face a developer console rather than a signup page. Tracks also cap at three minutes, and RealTime is still labeled experimental.
Play one prompt through and change the key while it is still running. Fade it out only when the idea has landed.