Stability Ai’s startup has unveiled the latest Stable Audio Open Small II model, specifically designed to produce stereo sound directly on mobile devices. The developers claim that this audio generator is the fastest on the market, capable of creating 11 seconds of audio in less than 8 seconds on a smartphone.
This model is the result of a collaboration with ARM, a company renowned for its chips used in smartphones and tablets. Unlike competitors such as Suno and UDIO, which rely on cloud processing, Stable Audio Open Small can operate offline, providing new possibilities for musicians, game developers, and mobile applications.
With 341 million parameters, the model is optimized for ARM processors and is ideal for producing short musical segments like jingles, instrumental parts, and sound effects. Stability Ai highlights that the model was trained using only open and free libraries such as the Free Music Archive and Freesound, minimizing copyright risks. In contrast to Suno and Udio, which are associated with secure content usage, Stable Audio Open Small is legally safe.
Nevertheless, the model has limitations: it only comprehends English requests and cannot generate realistic vocals or high-quality tracks. Stability also cautions that the output quality is influenced by genre and musical style, as most of the training was based on Western music.
The documentation on Hugging Face provides technical insights and constraints of the model, along with usage terms. It is free for research, personal use, and companies with annual revenue below $1 million; others require a paid Stability license.
Stable Audio Open Small marks a significant advancement by Stability AI amidst a challenging reorganization phase. Following financial difficulties and a change in leadership, the company revamped its board with the addition of director James Cameron and introduced a series of new generative products.
For demonstrations and audio samples,