Despite a drastic shift in its leadership recently, Stability AI has just released its latest and greatest AI model for music generation called Stable Audio 2. As the name says, this is an upgrade over the original Stable Audio with more powerful features.

The firm highlights that this updated version redefines the standards for music produced by artificial intelligence, enabling the creation of harmonious tracks of up to three minutes with studio-grade audio at a resolution of 44.1 kHz in stereo.

A notable enhancement that Stable Audio 2 brings is that users can not only create AI audio from text prompts but also from their own audio tracks. It has support for a new audio-to-audio feature that allows anyone to upload their own audio files and create fully produced samples.

Furthermore, Stability AI has joined hands with Audible Magic and uses its Audio Content Recognition (ACR) tech to safeguard the intellectual property rights of music creators. This tech is able to match generated content in real-time to prevent copyright infringement.

Creating sound and audio effects from scratch has also become easier with Stable Audio 2 since it offers enhanced options such as keyboard typing to crowd noise and city ambiance. The latest update to Stable Audio 2 introduces an innovative style transfer capability, offering users the flexibility to tailor their fresh compositions or uploaded audio files to fit the aesthetic of their projects seamlessly.

Stable Audio 2 can be accessed for free on its dedicated Stable Audio website as well as a 24/7 live stream on YouTube called Stable Radio. This can be found on the official Stable Audio YouTube channel. This live stream showcases a curated selection of tracks, each uniquely generated by the AI-powered tool. The launch of Stable Audio 2 follows its predecessor, Stable Audio 1, which made its debut in September 2023.

Similar to its predecessor, Stable Audio 2 has also been trained on data from the AudioSparx music library which has more than 800,000 audio files that are divided into a variety of instrument tracks, sound effects, music, and more, each accompanied by relevant text metadata.

Stability AI has emphasized that artists featured within the AudioSparx catalog were provided the option to exclude their works from the AI’s training dataset, ensuring respect for creator consent.