- Amuse 3.1 enables the creation of AI-generated images directly and locally on laptops with AMD Ryzen AI processors with XDNA 2 NPU.
- It uses the Stable Diffusion 3 Medium FP16/BF16 model, optimized to improve quality and reduce memory consumption by up to 9 GB on computers with 24 GB of RAM.
- The system delivers a final resolution of 4MP (2048x2048) thanks to a two-stage pipeline that performs integrated upscaling on the NPU.
- The collaboration between AMD and Stability AI drives advanced local AI capabilities, marking a shift from cloud processing and improving privacy and efficiency.
AMD and Stability AI collaboration takes a key step in the integration of generative artificial intelligence into personal computers. Until now, most advances in AI for image creation have relied on the cloud, but the new version of Amuse 3.1 transforms this reality by allowing the completely offline image generation and rescaling on laptops equipped with AMD Ryzen AI processors and its XDNA 2 NPU.
This technology It is presented as a solution designed for professionals, creators and designers who seek to perform tasks of generation of high-quality images without depending on external services nor face the limitations of an internet connection. It also promotes privacy and immediacy, since the entire process takes place on the user's own computer.
A truly local generative AI
With Amuse 3.1, laptops that incorporate AMD Ryzen AI 300 or AI MAX+ can access the creation of visual content using the new model Stable Diffusion 3 Medium in FP16 or BF16 format. This optimization allows the local image generation reach a resolution of X (4MP) from an initial base of 1024 x 1024 pixels, all thanks to a two-phase pipeline that first generates the image and then scales it, entirely on the NPU.
The system significantly reduces memory consumption compared to previous implementations. Although It is recommended to have 24 GB of RAM, the model itself only uses about 9 GB during generation, which allows for this experience on premium segment laptops, without needing to exceed 32 GB of memory.
Furthermore, the integration of the FP16/BF16 model provides a balance between accuracy, color quality and performance, falling somewhere between traditional FP16 calculations and the high performance of INT8, yet avoiding the quality loss common with over-quantization. Users can achieve detailed, customizable visual results, even suitable for professional printing.
Benefits and requirements for creative users
The new Amuse 3.1 is aimed primarily at those who need Create stock images, graphic design resources, or brand-specific visuals.. The user can start the generation by simple text prompts, which are particularly sensitive to structure, composition, and written details. Small changes in the precision of the prompt have a significant impact on the final result.
The supports advanced options, such as negative prompts to exclude unwanted elements or specific configurations to obtain images more faithful to the requirements of each project. Also It facilitates rapid iteration, allowing different variants to be generated by changing seeds or parameters. and select the most appropriate option for each need.
To access these functions, you just need a laptop with Ryzen AI 300/AI MAX+, XDNA 2 NPU with at least 50 TOPS and a minimum of 24GB of RAM. Amuse 3.1 is available as a beta download from Tensorstack, and it's recommended to install the latest AMD Adrenalin drivers to optimize NPU utilization. Within the application, the user must activate HQ mode and enable the 'XDNA 2 Stable Diffusion Offload' option for the entire process to run locally.
Implications for the AI and privacy industry
AMD's advancement in Stability AI marks a clear position compared to other manufacturers, such as NVIDIA and Intel, which are also focused on developing local generative AI solutions. AMD's approach focuses on decentralize content generation, boosting speed, efficiency and privacy, and introducing a new standard in the industry: the possibility of perform complex AI tasks, such as creating high-quality images, without relying on remote servers or cloud connections.
La Stability AI Community License regulates the use of the model, allowing its free exploitation for individuals, research projects, and small businesses with a volume of less than $1 million annually. For larger organizations, An Enterprise model is available under specific license.
This movement occurs in a context marked by legal debates about copyright and the use of data to train models IA, challenges that Stability AI addresses by offering greater transparency and controlled license management, while diversifying its catalog of generative tools with applications in image, video, and audio.
AMD and Stability AI demonstrate with Amuse 3.1 that the Advanced AI image generation is now viable in local environments, integrating efficiency, quality, and control over the creative process, enabling greater technological independence and data security.
I am a technology enthusiast who has turned his "geek" interests into a profession. I have spent more than 10 years of my life using cutting-edge technology and tinkering with all kinds of programs out of pure curiosity. Now I have specialized in computer technology and video games. This is because for more than 5 years I have been writing for various websites on technology and video games, creating articles that seek to give you the information you need in a language that is understandable to everyone.
If you have any questions, my knowledge ranges from everything related to the Windows operating system as well as Android for mobile phones. And my commitment is to you, I am always willing to spend a few minutes and help you resolve any questions you may have in this internet world.