AI startup Stability AI has introduced a new audio-generating model named Stable Audio Open Small, which the company claims is both the fastest in its class and efficient enough to operate directly on smartphones. This release is the result of a partnership between Stability AI and chip manufacturer Arm, a leading provider of processors for mobile devices such as phones and tablets.
Unlike many existing AI audio-generation tools like Suno and Udio, which often rely on cloud-based processing, Stable Audio Open Small is designed explicitly to function offline on portable hardware. Stability AI emphasizes that the entire training dataset for Stable Audio Open Small consists of royalty-free audio, sourced from publicly available libraries Free Music Archive and Freesound. This approach differs significantly from competitors whose training data has sometimes included copyrighted material, exposing potential intellectual property risks.
The new model is relatively compact, comprising 341 million parameters—this term refers to the internal elements that determine the AI’s behavior and responsiveness. Built specifically for Arm CPUs, it is optimized to generate short audio segments swiftly, producing up to 11 seconds of audio output in less than eight seconds of processing time on a typical smartphone, according to Stability AI.
Despite these significant developments, the company notes several limitations of the Stable Audio Open Small model. It supports prompts written exclusively in English, and Stability AI explicitly states in its documentation that users should not expect realistic human vocals or fully produced musical pieces. Moreover, due to its predominantly Western-focused training data, the model’s quality and accuracy vary across different genres and musical styles, potentially limiting its versatility.
The model is free to use for personal projects, academic research, and businesses generating less than $1 million in annual revenue. Organizations or developers that exceed this revenue threshold must acquire a paid enterprise license to utilize the technology commercially.
This product release arrives at a pivotal time for Stability AI, the creator behind the widely recognized image generation tool Stable Diffusion. The company recently faced significant challenges, including financial instability and internal crises precipitated by internal mismanagement and unsuccessful business collaborations. Over the past year, Stability AI has attracted new investment capital from high-profile backers, restructured leadership by appointing a new CEO, and even welcomed notable figures such as filmmaker James Cameron to its board.
Alongside its renewed management direction, Stability AI has focused on expanding its range of AI offerings, having released several improved versions of image generation models aimed at solidifying its position within the AI industry landscape.
