Stability has launched a new stereo audio generation model that has developed in collaboration with ARM to bring this technology directly to mobile devices.
Stable Audio Open Small is a “smaller and smaller” version of Stable Audio Open, launched last year with 1,100 million parameters to generate audio samples and sound effects of up to 47 seconds, such as battery rhythms or environmental sounds, based on text indications.
The new model is optimized for ARM technology, present in 99 percent of the ‘smartphones’ globally, as Stability has pointed out. It has 341 million parameters and is designed to generate stereo audio in less than eight seconds from text.
With this model you can generate audio samples, sound effects and production elements, such as battery loops, environmental instruments and textures riffs, as the company explains in a statement.
Stable Audio Open Small has been shared under an open source modality. The pesos of the model can be found in Hugging Face and the Code, in Github.