Stability AI, an innovative AI startup, has introduced Stable Audio Open Small, a cutting-edge audio-generating model touted as the quickest of its kind, capable of operating efficiently on smartphones. This development results from a partnership with Arm, the renowned chipmaker responsible for processors in many mobile devices. Unlike existing audio-generating applications that typically depend on cloud processing—like Suno and Udio—Stable Audio Open Small allows offline functionality, making it a convenient choice for users on the go.
The dataset used to train this model comprises royalty-free music from sources such as Free Music Archive and Freesound, thereby mitigating intellectual property risks often associated with copyrighted materials. Stable Audio Open Small features 341 million parameters and is specifically designed to work seamlessly with Arm CPUs. This makes it highly effective for generating brief audio snippets and sound effects, capable of producing up to 11 seconds of audio in less than eight seconds on a smartphone.
However, the model does come with certain limitations. Currently, it only supports prompts in English and has been noted for its inability to create realistic vocals or high-quality songs. Additionally, there are concerns regarding its performance across various musical styles, largely influenced by its Western-centric training data.
Developers should also be mindful of Stable Audio Open Small’s usage restrictions. While the model is available at no charge for researchers, hobbyists, and firms earning under $1 million annually, those with revenues exceeding this threshold will need to acquire Stability AI’s enterprise license.
Stability AI made headlines previously with their popular Stable Diffusion image generation model, but faced difficulties under former CEO Emad Mostaque’s leadership, including financial struggles that led to staff departures and cancelled partnerships. In response, the company has appointed a new CEO and brought on notable figures like James Cameron to its board of directors to help turn things around. Recently, it has also introduced new image generation models, signalling a renewed focus and direction for the firm.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
