Stable Audio

2024: Product Announcement

On April 3, 2024, Stability AI introduced the Stable Audio 2.0 artificial intelligence model, which is said to set new standards for the generation of audio materials. In particular, this neural network is capable of creating full-fledged tracks lasting up to three minutes.

Stable Audio 2.0 can generate original content based on user-uploaded audio recordings according to natural language prompts. It is argued that the new network differs from other similar AI models in that it creates compositions with a full structure - introduction, development and ending. At the same time, stereo effects are supported, and the sampling rate reaches 44.1 kHz. It is noted that the neural network is capable of generating ambient sounds, for example, crowd noise or tapping on the keyboard. You can also generate materials based on text hints only.

Stability.AI

The presented AI model is based on the Stable Audio 1.0 neural network, which debuted in September 2023. Both versions were trained on data from the AudioSparx music library, which contains more than 800 thousand audio files, including music, sound effects and sounds of individual instruments, as well as related text metadata. At the same time, all performers are given the opportunity to prohibit the use of their works to teach AI models.

The rules of the service prohibit the use of the Stable Audio 2.0 neural network to generate tracks based on copyrighted audio materials. Advanced content recognition tools are used to meet this requirement and prevent violations. The new AI model is completely free: it is available through the Stability AI website, as well as through the program interface (API).^[1]

Notes