The stability of AI startups has published an open Small stable audio model, an AI model generating “stereo” audio which, according to the company, is the fastest on the market – and sufficiently effective to operate on smartphones.
Stable Audio Open Small is the fruit of a collaboration between AI and ARM stability, the flea manufacturer that produces many processors inside tablets, phones and other mobile devices. Although a number of applications fueled by AI can generate audio, like Suno and Udio, most rely on the treatment of clouds, which means that they cannot be used offline.
Stability also claims that the stable Audio Open Small training set is entirely composed of songs from the free music library and free -free music. It is as opposed to the above -mentioned Suno and Udio training sets, which contain content protected by copyright, posing an IP risk.
Stable Audio Open Small is 341 million size settings and optimized to operate on ARM processors. (The parameters, sometimes called weight, are the internal components of a model that guide its behavior.) Designed to quickly generate short audio samples and sound effects (for example, drum and instruments riffs), an open stable audio can produce up to 11 seconds of audio on a smartphone in less than 8 seconds, claims the AI stability.
Here is a sample generated by Stable Audio Open Small:
And here is another:
The model is not without its limits. Stable Audio Open Small only supports the invites written in English, and notes of stability in its documentation that the model cannot generate realistic votes or high quality songs. The model does not work well through musical styles, prevents stability – a consequence of its western biased training data.
In another potential wrinkle for developers, an open Small stable audio has somewhat restrictive use terms. It is free for researchers, amateurs and businesses with less than a million dollars in annual income, but developers and organizations earning more than a million dollars must pay the business license.
Stability, the company besieged behind the stable broadcast of the popular image generation model, collected new species last year while investors, including Eric Schmidt and the founder of Napster, Sean Parker, sought to reverse the share of the company. Emad Mostaque, co-founder and ex-CEO of stability, would have poorly managed stability in financial ruin, which led the staff to resign, a partnership with Canva to pass and investors to worry about the prospects of the company.
In recent months, stability has hired a new CEO, appointed Titanic James Cameron’s director to his board of directors and has published several new models of image generation.
