Stability AI, the corporate behind AI-powered picture generator Secure Diffusion, has launched Secure Audio Open, an open supply mannequin for producing quick audio samples, sound results and manufacturing parts utilizing textual content prompts.
The brand new mannequin was skilled on audio knowledge from free music libraries Freesound and the Free Music Archive. “This allowed us to create an open audio mannequin whereas respecting creator rights,” says Stability AI. The corporate provides that Secure Audio Open’s specialised coaching makes it splendid for creating drum beats, instrument riffs, ambient sounds, foley recordings and different audio samples for music manufacturing and sound design.
Customers can generate as much as 47 seconds of audio knowledge by inputting textual content descriptions like “heat arpeggios on an analog synthesizer with a steadily rising filter cutoff and a reverb tail” and “rock beat performed in a handled studio, session drumming on an acoustic equipment”.
One key benefit of the open supply launch is that customers can fine-tune the mannequin on their very own customized audio knowledge. For instance, a drummer might fine-tune on samples of their very own drum recordings to generate new beats.
That mentioned, whereas Secure Audio Open can generate quick musical clips, it’s not optimised for full songs, melodies or vocals in contrast to the corporate’s flagship Secure Audio service. The latter is ready to produce tracks with coherent musical construction as much as three minutes in size, and presents superior capabilities like audio-to-audio technology and coherent multi-part musical compositions.
In keeping with Stability AI, the open supply mannequin “gives a glimpse into generative AI for sound design whereas prioritising accountable growth alongside artistic communities.”
The corporate’s newest deal with ‘accountable audio technology’ follows the high-profile exit of its VP of generative audio, Ed Newton-Rex, final November, who give up resulting from disagreements with the agency over what constitutes “truthful use” of copyrighted works.
The previous govt mentioned he disagreed “with the corporate’s opinion that coaching generative AI fashions on copyrighted works.” Newton-Rex additionally advised the BBC that he thought it was “exploitative” for builders to make use of artistic work with out consent – a stance he claimed many AI corporations, together with Stability AI, would beg to vary.
Get the most recent information, opinions and tutorials to your inbox.
