← Constellation
Audio & Music/Music Generation

Riffusion

AI music from text descriptions using spectrograms

Visit riffusion.com

External link. Not endorsed — curated for usefulness.

What is Riffusion?

Riffusion is an AI music generation tool that creates original music from text descriptions using spectrogram-based synthesis, made by Google. Users input descriptions of musical styles, moods, instruments, or genres, and the system generates full-length songs with vocals and dynamic arrangements. The tool uses Google's Lyria 3 frontier music model to produce compositions with rich musicality and control over fine details.

The platform operates as a browser-based studio with a chat interface that mimics working with a music producer. Users can compose complete tracks, adjust audio stems, apply effects, and refine individual elements through an intuitive dashboard. A built-in piano interface and preset collections (Electronica, R&B, and other styles) help users explore possibilities. The service integrates AI video generation via Google's Veo model, allowing users to create accompanying music videos by specifying visual aesthetics and characters. A "Vibe-code" feature enables developers to build custom audio plugins, music games, and DAWs using the underlying API.

Riffusion operates on a free-to-start model with daily credits for generation. No credit card is required for initial use. The platform includes playlist creation, music publishing, artist following, and community discovery features. Songs generated remain in users' personal libraries and can be shared or published publicly. The system learns individual user preferences over time, personalizing recommendations and generation style as more music is created on an account.

The tool targets musicians, producers, content creators, and hobbyists who want to generate royalty-free background music or explore composition without traditional music production experience. Riffusion competes with comparable AI music generators including Udio and Suno, which offer similar text-to-music capabilities with different model architectures and publishing features.