London-based startup Stability AI has just released Stable Audio, an AI-powered music generator capable of creating high-quality, commercial-ready music. This marks Stability AI’s renewed commitment to audio and its expansion into the field of AI music creation tools.
Stability AI has released Stable Audio, an AI-powered music generator that offers greater control and coherence than previous generative music tools. With its latent diffusion technique, Stable Audio can generate high-quality music tracks of up to 90 seconds. While questions about copyright and artistic compensation remain, Stable Audio represents Stability AI’s commitment to expanding its AI models into multimodal content creation.
A New Approach to AI Music Generation
Stable Audio is built on a roughly 1.2-billion-parameter model developed by Stability’s in-house audio team. This model, inspired by its predecessor Dance Diffusion, offers greater control over the content and length of synthesized audio. Unlike previous generative music tools, Stable Audio can generate longer audio tracks and provide users with more control using text prompts and desired durations.
The underlying technique used in Stable Audio is called latent diffusion, which gradually subtracts noise from a starting song made mostly of noise, moving it closer to the desired text description. This approach ensures that the generated music is coherent and melodic for up to 90 seconds, surpassing the limitations of other AI models that often devolve into random noise after a few seconds.
Unlocking the Potential of Generative AI
Ed Newton-Rex, VP of audio for Stability AI, emphasized that the company’s mission is to unlock humanity’s potential through foundational AI models across various content types. Starting with Stable Diffusion and now expanding to music, Stability believes that the future of generative AI lies in multimodality.
Generating a Wide Range of Music
Stable Audio is primarily designed to generate instrumental music and offers the most control and musicality in genres like EDM and ambient music. While it may lack some creativity, the music it produces sounds more coherent and melodic compared to other AI-generated music. Stability has provided samples across various genres, showcasing the capabilities of Stable Audio.
Monetization and Copyright Considerations
Stability AI customers can monetize their works generated using Stable Audio, though copyright protection is not provided. Stability AI partnered with commercial music library AudioSparx to train Stable Audio, ensuring that participating artists can share in the profits generated by the tool. However, concerns regarding copyright infringement and the use of artists’ work without proper compensation still persist.