Stability unveils Stable Audio Open, a new open-source model for music & audio generation

Stability AI, known for its image generator Stable Diffusion, has launched an AI model for audio generation called Stable Audio Open. This tool transforms text descriptions into audio recordings up to 47 seconds long (for now), using royalty-free music for training.

Stable Audio Open is based on the commercial Stable Audio tool, first released in September, with a more advanced version, Stable Audio 2.0, launched in April. Unlike the commercial version, Stable Audio Open is trained on 486,000 samples from free music libraries like Free Music Archive and Freesound, avoiding copyrighted material. While advertised as open-source, Stable Audio Open is available under a noncommercial research community agreement license, restricting commercial use of the generated sounds.

Targeting musicians and filmmakers, the tool is ideal for generating drum beats, instrument riffs, ambient sounds, and production elements. It can also edit existing songs or apply new styles, such as converting a piece to smooth jazz. Users can fine-tune the model with their own audio to create sounds in their unique style. However, it is not intended for full songs or vocals, and for longer music, users should consider the premium Stable Audio 2.0 service or other alternatives like Suno AI or Udio.