AMD and Stability AI said on Monday that the companies have ported stable Amuse AI art generators from stable diffusion 3.0, improving image generation and even allowing text to be run exclusively on NPUs.
With most jaw-dropping advances in AI art from the cloud (such as Google Veo 3), the new AI art models running locally on AMD’s PC processors don’t seem to deserve coverage. But it’s a step forward for AMD and is lagging behind some of the advancements in AI art from Intel.
The new model will be part of Amuse 3.1, which can be downloaded from Stability AI. Download stable spreading media models via the Amuse app. I tried this in April when Amuse started adding text to videos. The new model uses what is known as BF16. This allows for a wider range of responses, but reduces accuracy. It also works better in a way than some of the results that were able to be produced in April.
Stable diffusion media is already available by Amuse, but previously required a GPU, limiting the appeal of mass. Now, which one can I choose? As more and more laptops start to add NPUs, this opens up local image generation to more viewers.
Mark Hackman/Foundry
AMD provided a press preview last week and tried it out using the MSI Prestige A16 AI+. Each image (with 20 steps) took about 90 seconds to generate. It’s still far behind in the cloud, but image generation on local PCs is unlimited.
You can know whether Amuse still makes mistakes and sees well that it is not created by human hands. But it is steadily improving. Where are you in a year or two?
Mark Hackman / Foundry / Amuse
Fix: The subheadline of this story incorrectly said that you need a laptop with 32GB of RAM. This model works on a Ryzen laptop with 24GB of RAM.
