Stability AI and Arm Collaborate to Release Stable Audio Open Small for On-Device Audio Generation

Stability AI has partnered with Arm to open-source Stable Audio Open Small, a lightweight text-to-audio generation model with 341 million parameters. The model is optimized for Arm CPUs, enabling near-real-time audio generation in on-device environments such as smartphones.

Designed using the Arm KleidiAI library, this model offers the following technical strengths:

• Lightweight: A much smaller 341M parameter size compared to the existing Stable Audio Open (1.1B parameters).
• Fast generation: High-speed generation capable of producing 11 seconds of audio in under 8 seconds on a smartphone.
• Efficiency: Runs entirely on Arm CPUs, allowing execution on edge devices at low cost without the need for separate high-performance hardware.

The model is optimized for generating short audio samples and sound effects such as drum loops, foley, instrument riffs, and ambient textures. It is particularly well-suited for mobile and edge device deployment where real-time responsiveness is critical.

Stable Audio Open Small is available free of charge for both commercial and non-commercial use under the Stability AI Community License. Model weights are available on Hugging Face, and the code can be found on GitHub.