Stability AI has launched Stable Audio, a generative AI platform that creates audio based on textual cues.
Here's What We Know
Stable Audio uses a diffusion model trained on 800,000 audio files with music, sound effects, and single-voice singing. Textual metadata from AudioSparx was also used.
Stability AI says it has permission to use the copyrighted material.
A distinctive feature of Stable Audio is its ability to generate tracks of different lengths. To do this, the timing information in the training data was analysed.
The platform has a free version with a limit of 20 tracks up to 45 seconds per month. The paid Professional plan for $11.99 allows you to create up to 500 tracks of 90 seconds each.
Users of the free version cannot use the generated audio for commercial purposes.
Source: Stability AI