Stability AI unveils Stable Audio 2 for AI-generated music and audio effects



summary
Summary

Stability AI has released Stable Audio 2, an advanced version of its AI-powered music generation tool.

According to the company, this new model sets a new benchmark for AI-generated music by composing coherent songs up to three minutes long in studio quality at 44.1 kHz stereo.

One of the key features of Stable Audio 2 is the audio-to-audio function, which allows users to upload their own audio files and create fully produced samples. To protect the copyrights of music creators, Stability AI has partnered with Audible Magic and uses their Audio Content Recognition (ACR) technology to match content in real time and prevent copyright infringement.

The model also offers enhanced options for creating sound and audio effects, ranging from keyboard typing to crowd noise and city ambience. A style transfer feature has also been added, allowing users to match newly created or uploaded audio files to the desired style of their project.

Ad

Ad

Stable Audio 2 is now available for free on the Stable Audio website, and a 24/7 live stream called “Stable Radio” can be found on the Stable Audio YouTube channel, featuring tracks exclusively generated by the AI tool. The previous version, Stable Audio 1, was launched in September 2023.

Training data and ethical concerns

Like its predecessor, Stable Audio 2 was trained solely on data from the AudioSparx music library, which consists of over 800,000 audio files with music, sound effects, and individual instrument tracks, along with corresponding text metadata. Stability AI stated that all AudioSparx artists had the opportunity to opt out of the training process.

Ed Newton-Rex, former head of Stability AI’s audio team, disagreed with the company’s view that training generative AI models on copyrighted works could be considered “fair use.” He emphasized that training AI models without the permission of creators is morally wrong and potentially undermines their livelihoods, and urged other AI company employees to advocate for fairer treatment of creators.

This issue is currently in the spotlight, as a group of music creators, including well-known artists like Billie Eilish and Katy Perry, recently wrote an open letter to AI companies, demanding them to stop using AI that violates and devalues the rights of human artists, calling generative AI models an “attack on human creativity.”

Right now, it looks like everyone is a loser. Stability AI itself is facing financial difficulties, as training generative AI models has not yet proven profitable for the AI startup, leading to the resignation of CEO Emad Mostaque due to alleged mismanagement.

Recommendation

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top