Stable Audio 2.0
Stable Audio 2.0 makes coherent three-minute tracks at 44.1 kHz stereo and adds audio-to-audio, trained only on licensed AudioSparx data.
Early example of a licensed-data music model with opt-out and creator compensation; audio-to-audio lets users transform uploaded samples with prompts. Stable Audio Open (2024-06-05) followed as an open model for samples up to 47 seconds.
- Date
- Wednesday, 3 April 2024
- Lab
- Stability AI
- Kind
- model
- Access
- app only
Stable Audio Open is open source (up to 47 s).
Sources
This record was checked against its sources on 6 October 2026. How we check