Description
Stable Audio 3 Medium is part of the Stable Audio 3 family, which includes small, medium, and large models for audio generation and editing. These models utilize fast latent diffusion techniques to produce audio of varying lengths, making them ideal for generating short sounds without the overhead of full-length audio production. The models are built on a semantic-acoustic autoencoder that compresses audio into a compact latent space, ensuring efficient generation while maintaining audio fidelity and semantic structure.
Read More
The Stable Audio 3 models undergo adversarial post-training to enhance inference speed and generation quality, reducing the number of inference steps required while improving fidelity and adherence to prompts. These models are trained on a dataset comprising licensed and Creative Commons audio recordings, allowing them to generate music and sounds quickly on consumer-grade hardware, such as an H200 GPU or a MacBook Pro M4.
For commercial use, users must refer to Stability AI's licensing terms. The models can be integrated using the stable-audio-3 inference and fine-tuning library or the stable-audio-tools research library. The models are designed for users seeking efficient audio generation and editing capabilities, particularly those interested in music and sound production.
- CategoryAI Music Generators
- Pricing ModelRefer to Stability AI
- DeploymentCloud
- PlatformsWeb
Stable Audio 3 Medium Highlights
Latent diffusion model
Variable-length audio generation
Audio inpainting support
Semantic-acoustic autoencoder
Adversarial post-training
Licensed and Creative Commons data
Fast inference on consumer hardware
Integration with stable-audio-tools
Getting Started with Stable Audio 3 Medium
Access model: Visit Hugging Face repository
Authenticate: Agree to terms and conditions
Set up environment: Prepare hardware and libraries
Integrate via API: Use stable-audio-3 library
Optimise: Fine-tune model for specific needs
Stable Audio 3 Medium's Use Cases
- Music Production
- Sound Design
- Podcast Editing
- Audio Restoration
- Creative Commons Projects
