Stability AI releases audio generation models that can run on smartphones

AI Launch Stability AI released a stable audio open small, the company claims to be the fastest “stereo” audio-generated AI model on the market and is efficient enough to run a smartphone.

Stable audio opening is the fruit of a collaboration between stable AI and ARM, a chip manufacturer that produces many processors in tablets, phones and other mobile devices. While many AI-powered applications can generate audio, such as Suno and Udio, most rely on cloud processing, which means they cannot be used offline.

Stability also claims that the stable audio open training suite consists entirely of songs from free free audio libraries, free music archives and free music. This is contrary to the above-mentioned Suno and Udio training sets, which reportedly contain copyrighted content, poses IP risks.

The stable audio open volume is 341 million large and optimized to run on ARM CPU. (The parameter sometimes called weights is an internal component that guides its behavior.) Designed to quickly generate short audio samples and sound effects (e.g., drums and drums and instrument improvisation), stable audio can produce up to 11 seconds of audio on a smartphone in less than 8 seconds.

Here is a small sample opened by stable audio:

Here is another one:

This model is not without limitations. The small support for stable audio on the open supports only tips written in English and notes in its documentation that the model cannot produce realistic vocals or high-quality songs. Stability warning, the model also performs poorly in musical style, a result of its western biased training data.

In another potential wrinkle in Devs, the small terms of use for stable audio opening are somewhat limited. It can be used for free for researchers, hobbyists and businesses that earn less than $1 million a year, but developers and organizations that earn more than $1 million must pay for stable corporate licenses.

Stability is the troubled company behind the steady spread of popular image generation models, which last year as investors (including Eric Schhmidt and Napster founder Sean Parker) proposed new cash in an attempt to change the business. Stable co-founder and former CEO Emad Mostaque reportedly made the employees resign, partnerships with Canva, and focus on the company's prospects.

Over the past few months, Stability has hired a new CEO, appointed Titanic director James Cameron to the board, and released several new image generation models.