Stability AI has made Stable Audio free and publicly available, unlike similar models from companies like Meta and Google. The model is trained on data from AudioSparx, which has a library of over 800,000 audio files. To avoid copyright issues, Stability AI has partnered with Audible Magic to use its content recognition technology to track and block copyrighted material from entering the platform. Despite improvements, Stable Audio 2.0 still struggles with creating songs that sound authentic and have discernible language.
Key takeaways:
- Stability AI has released Stable Audio 2.0, an improved version of its audio generation model that allows users to upload their own audio samples and transform them into AI-generated songs.
- Unlike its predecessor, Stable Audio 2.0 offers a full three-minute sound clip and the ability to create songs with an intro, progression, and outro. However, the AI-generated songs still lack the soul and feel of human-created music.
- All uploaded audio must be copyright-free. Stability AI has partnered with Audible Magic to use its content recognition technology to track and block copyrighted material from entering the platform.
- Stable Audio is trained on data from AudioSparx, which has a library of over 800,000 audio files. Artists under AudioSparx were given the option to opt out of their material being used to train the model.