May 15, 02:40 PM

Stability AI Launches 341M-Parameter Stable Audio Open Small Model Running on Arm CPUs for Fast Mobile Audio Generation

Stability AI, in collaboration with Arm and researchers from UC San Diego, has launched Stable Audio Open Small, a compact text-to-audio generation model with 341 million parameters. This model is designed to run efficiently on Arm CPUs, which power approximately 99% of smartphones, enabling users to generate sound effects directly on their mobile devices without the need for GPUs. The model can produce 11 seconds of audio in under 8 seconds, offering fast and royalty-free sound generation. The release also features a novel adversarial relativistic-contrastive (ARC) post-training technique that enhances the speed, diversity, and efficiency of text-to-audio generation across various devices. Stable Audio Open Small is free to use and supports quick fine-tuning, marking a significant advancement in mobile audio AI capabilities.

#Stability AI #Arm #UC San Diego #Stable Audio Open Small

Written with ChatGPT (GPT-4).

Stability AI Launches 341M-Parameter Stable Audio Open Small Model Running on Arm CPUs for Fast Mobile Audio Generation

Sources

Additional media