Amuse 2.2 Beta is now available for AMD Ryzen™ AI 300 Series processors and Radeon™ 7000 Series graphic users and can be downloaded from the Amuse AI website. Amuse Beta is a dependency free, AI image generation studio optimized for AMD hardware. The new 2.2 Beta version brings support for Stable Diffusion 3.5 (SD 3.5) family of models by Stability AI: including Stable Diffusion 3.5 Medium, Stable Diffusion 3.5 Large and Stable Diffusion 3.5 Large Turbo. It also includes an image quality enhancement for AMD Ryzen™ AI family of products.
Stable Diffusion 3.5 Support
Stable Diffusion 3.5 family of models introduce a 8.1 billion parameter model called Stable Diffusion 3.5 Large. This base model provides a powerful addition to the Stable Diffusion family and presents state of the art image capability for 1 MP (~1024×1024) images. Stable Diffusion 3.5 Large Turbo is a distilled version of the same model which can generate an image in just four steps with a slight quality trade off.
Stable Diffusion 3.5 Medium is another new model with 2.5 billion parameters and can generate images all the way up to 2 Mega Pixels (~1536×1536). Since it is a smaller model, it also has a smaller memory footprint and wider hardware support. Both the SD 3.5 Large and SD 3.5 Medium models are finetuneable and feature improved text support and prompt adherence.
All the Stable Diffusion 3.5 models are now available for AMD users starting with the Amuse 2.2 Beta release. Amuse allows partial offload of larger models that may not fit inside the vRAM of the device using a “minimum” mode, which sequentially loads the different parts of an image generation model. Performance in this mode is affected by your disk read/write speeds and RAM speeds.
Significant image quality updates have also been rolled out for AMD Ryzen™ AI 300 Series processors. Highly efficient architectures and checkpoints like Stable Diffusion Lightning are now fully supported and will be the default “HQ” mode while running Amuse in EZ Mode. This update brings a significant uplift to local AI image generation on thin and light systems while on-the-go and completely offline.
The new HQ model in Amuse 2.2 Beta can deliver high quality images with significantly increase prompt adherence and visual fidelity. Images are now more detailed and coherent. The HQ mode now also features a ”boost quality” option as well, which will increase image quality even further in most cases at the cost of increasing the generation time. To try out this new HQ model, download Amuse on an AMD Ryzen™ AI 300 Series laptop and change the quality slider to “HQ”.
The AMD XDNA™ architecture powers Super Resolution features has also been optimized and is now capable of end-to-end upscaling much faster than before. Combined with the AI Filters and AI Design tab which house powerful AI assistants, Amuse is a very easy to use AI toolkit that requires no technical knowledge and auto configures.