1. StabilityAI has released Stable Audio 2.0, an AI music generation tool that allows users to create three-minute tracks using natural language processing prompts.
2. The new version includes audio-to-audio support, where users can create “fully produced samples” by uploading an audio file to the platform.
3. The company has reinforced its copyright protection policies, with content recognition technology to prevent copyright infringement and the option for musicians to opt out of having their works used for AI model training.
StabilityAI has launched Stable Audio 2.0, an AI music generation tool that allows users to create three-minute tracks at 44.1 kHz stereo by inputting a natural language processing prompt. The tool now offers longer tracks, audio-to-audio support, and a greater focus on copyright protection for creators. Users can now also generate “fully produced samples” by uploading an audio file to the platform. The tool is trained on AudioSparx’s vast audio file library but creators have the option to opt out if they do not want their works used for AI model training.
Former VP of audio, Ed Newton-Rex, recently resigned from StabilityAI due to disagreements with the company’s opinion on training generative AI models on copyrighted works. He criticized the company’s approach to upholding creator’s rights and urged creators to voice their concerns to prevent exploitation by tech companies. Despite this controversy, Stable Audio 2.0 has been upgraded with longer tracks, audio-to-audio support, and a new architecture to enhance performance and generate full tracks with coherent structures. The tool incorporates compressed autoencoder and diffusion transformer technologies to recognize and reproduce essential structures for high-quality musical compositions. Stable Audio 2.0 is available for free use immediately.