Stability AI and Arm Launch Stable Audio Open Small for On-Device Audio Generation
May 14, 2025
Key Highlights
- Stable Audio Open Small is a 341 million parameter text-to-audio model optimized to run entirely on Arm CPUs.
- It generates up to 11 seconds of audio on a smartphone in under 8 seconds.
- Developers can access the model weights, research paper, and code for free under the Stability AI Community License.
- An Arm Learning Path is available for hands-on guidance deploying the model on Arm hardware.
Bringing Generative Audio to Smartphones
Stable Audio Open Small is the product of a collaboration between Stability AI and Arm, whose technology powers 99% of smartphones worldwide. This compact version of the Stable Audio Open model maintains output quality and prompt accuracy while being smaller and faster.
Following the demonstration of AI-generated audio running on Arm CPUs at Mobile World Congress, the model is now accessible for developers to deploy on mobile devices. This makes it possible to create text-based audio samples directly on smartphones without relying on cloud infrastructure.
Technical Advantages
- Lightweight: With 341 million parameters, it's significantly smaller than the original Stable Audio Open model, which has 1.1 billion parameters.
- Fast: Audio generation takes less than 8 seconds on a mobile device, allowing quick creation and fine-tuning.
- Efficient: Utilizing Arm’s KleidiAI libraries, the model runs efficiently on-device, reducing compute costs and removing dependency on heavy hardware.
Best Use Cases
Stable Audio Open Small is optimized for generating short audio clips such as drum loops, foley sounds, instrument riffs, and ambient textures. Its speed and compact size make it ideal for real-time applications on Arm-powered smartphones and edge devices.
As AI-generated creative tasks move to edge devices, smaller models like this help allocate processing resources effectively. Organizations can choose model sizes based on their needs, whether producing short sound effects or longer audio content.
Getting Started with Stable Audio Open Small
The model is freely available for both commercial and non-commercial use under the Stability AI Community License. You can:
- Read the research paper on arXiv.
- Download the model weights on Hugging Face.
- Access the code repository on GitHub.
For developers looking to deploy the model on Arm hardware, the Arm Learning Path offers step-by-step guidance. The Arm Community Blog provides detailed insights into the optimizations enabling efficient on-device performance.
Stay informed about updates and community discussions by following relevant channels on social media platforms and joining Discord groups focused on AI audio generation.
For those interested in expanding their AI skills, explore comprehensive courses at Complete AI Training.
Your membership also unlocks:
