Lovable: Lyria 3 Music Generation for App AI Features
Lovable added two Google Lyria 3 music models to the AI features available inside apps. Lyria 3 Clip Preview generates clips of about 30 seconds, and Lyria 3 Pro Preview produces full tracks up to 184 seconds. Apps can generate music from text prompts or images, with instrumental or vocal output, and credits are charged for each completed track.
Key Takeaways
- Two Google models are now available for app AI features: Lyria 3 Clip Preview and Lyria 3 Pro Preview.
- Clip Preview makes roughly 30-second clips, which fits short loops, jingles and quick previews.
- Pro Preview makes tracks up to 184 seconds, long enough for a complete song or extended background music.
- Text prompts and images can both drive generation, so apps can compose music from a description or from a picture.
- Instrumental or vocal output can be requested, which widens the range of apps that can use it.
- Credits are charged per successful track, so builders should account for the cost when they ship music features to users.
Music generation arrives in Lovable apps
Lovable added two Google music models to the AI features that builders can put inside their apps: Lyria 3 Clip Preview and Lyria 3 Pro Preview. Apps built on Lovable can now generate original music on demand instead of only text, images and other media.
Two models for two lengths
Lyria 3 Clip Preview generates short clips of approximately 30 seconds. It suits jingles, sound effects, loops and quick previews. Lyria 3 Pro Preview creates full tracks of up to 184 seconds, which suits complete songs and longer background music.
Prompts, images and vocals
Apps can request music from a text prompt or from an image. Each request can ask for an instrumental version or for vocals, so one app can serve a podcast creator who needs background music and a songwriting tool that needs sung lyrics.
How credits work
Music generation consumes credits for each successfully completed track. Builders adding the feature to an app should plan for that usage, since every generated track draws on credits.