On Monday, AMD and Stability AI announced their collaboration to integrate Stable Diffusion 3.0 Medium into Stability’s Amuse AI art generator. This enhancement enables improved image and text generation capabilities solely utilizing the NPU.
Although many recent strides in AI art have primarily emerged from cloud platforms, like Google Veo 3, this local implementation on AMD’s PC processors signifies a meaningful advancement for the company, which has traditionally trailed Intel in AI art technology. The newly updated model will be featured in Amuse 3.1, available for download from Stability AI.
Early previews indicate that the results from this version exhibit significantly enhanced quality compared to outputs generated back in April. Previously, users of Stable Diffusion Medium via Amuse were constrained by the requirement of a GPU, which limited its accessibility.
With the current updates, users can select between GPU and NPU, broadening the scope for local image generation as more laptops begin to incorporate NPUs. I had the opportunity to test Amuse 3.1 on an MSI Prestige A16 AI+ equipped with an AMD Ryzen AI 9 365 processor and an NPU offering 50 TOPS.
Each image, generated with 20 steps, took roughly 90 seconds. While this generation speed still lags behind cloud services, the advantage of local processing is the absence of usage limits.
While Amuse continues to make occasional errors—evident upon close examination—it is progressively improving. The potential future of local AI art generation raises intriguing questions about where technology may be in a year or two.
*Correction: The sub-headline of this story previously indicated a requirement for laptops with 32GB of RAM; however, this model operates with Ryzen laptops that have 24GB of RAM. *