Stability AI, in collaboration with Arm, has released Stable Audio Open Small, a powerful yet compact text-to-audio AI model designed to run directly on smartphones and other edge devices.

This new model has 341 million parameters and can generate up to 11 seconds of audio in under 8 seconds, all on an Arm-powered mobile phone with no extra hardware needed.

The release follows the companies’ earlier showcase at Mobile World Congress, where they demonstrated real-time AI-generated audio running on Arm CPUs. This new version makes the technology more accessible, allowing developers to explore audio generation without relying on large cloud-based systems.

Key Features of Stable Audio Open Small

Lightweight: Much smaller than the original (1.1B vs. 341M parameters).

Fast: Optimized for quick audio generation and easier fine-tuning.

Efficient: Designed to run using Arm’s KleidiAI tools, making it ideal for smartphones and edge devices.

The model is especially useful for generating sound effects, loops, foley sounds, and ambient music through simple text prompts. It's perfect for creators and developers building apps that need on-device, responsive audio generation.

Availability

• The model is free to use for both commercial and non-commercial projects under the Stability AI Community License.
• Developers can find the research paper on arXiv, model weights on Hugging Face, and code on GitHub.
• A new Arm Learning Path is also available to help developers deploy the model on Arm hardware.

This marks a big step in bringing AI-powered creativity to mobile and edge environments opening the door for new music apps, games, and audio tools that work fast, offline, and on-the-go.