Stability AI, the company behind open-source AI image models, is back with another model. We are jamming on music this time. The new audio model called Stable Audio 2.0³⁴ creates full music tracks with high-quality.

What’s going on here?

Stability AI launches Stable Audio 2.0 - an upgrade to their music generation model.

What does that mean?

The new model takes things to the next level with tracks that go up to 3 minutes long from a single prompt. Want more? It’s not just text-to-audio, Stable Audio 2.0 supports audio-to-audio conversion.

To drive that feature home, think of humming a beat in your voice and asking to model to generate that beat with a guitar. This new model has got you. It also emphasized sound effects better and can work with transferring styles.

Stability AI also says that the model was trained on a fully licensed dataset while keeping opt-out requests in mind. On that ground, they are allowing commercial use of this model too. But like their previous models, this isn’t open-source, so you can try it at stableaudio.com¹⁸ and opt for a commercial license if that’s your jam 😉.

Why should I care?

AI is expanding beyond just text and images. AI video, AI music, and AI speech are taking centre stage in 2024. Often there’s talk about deep fakes but Stable Audio works like a tool to create fun music that you can use in projects or just vibe on.

The question is do you really care if a human recorded the actual beats or they generated it using an AI model? PS: listen to this Ben’s Bites banger⁷ I made with Suno.