#stablediffusion3.5

Presenting Amuse 2.2 Beta: AMD Ryzen AI Image Quality Update and Stable Diffusion 3.5 Support.

Users of AMD Ryzen AI 300 Series CPUs and Radeon 7000 Series GPUs can download Amuse 2.2 Beta from the Amuse AI. Designed for AMD hardware, Amuse Beta is a dependency-free AI picture creation studio. Supporting Stability The current 2.2 Beta incorporates AI’s Stable Diffusion 3.5 (SD 3.5) family of models, including Medium, Large, and Large Turbo. Additionally, it features an improvement in picture quality for AMD’s Ryzen AI family of devices.

Support for Stable Diffusion 3.5

The 8.1 billion-parameter Stable Diffusion 3.5 Large model is part of the family. The base model provides advanced picture capabilities for 1 MP (~1024×1024) photos and is a powerful addition to the Stable Diffusion family. A condensed version of the same model, Stable Diffusion 3.5 Large Turbo, may produce a picture in just four stages at a little quality trade-off.

Another new model, Stable Diffusion 3.5 Medium, has 2.5 billion parameters and can produce pictures up to 2 Mega Pixels (~1536×1536). Being a smaller variant, it also supports a larger range of hardware and has a reduced memory footprint. Better text support and timely adherence are features of both the SD 3.5 Large and SD 3.5 Medium versions, which may be adjusted.

With the introduction of Amuse 2.2 Beta, AMD users may now access all of the Stable Diffusion 3.5 models. Amuse’s “minimum” mode, which loads the various components of an image generating model one after the other, enables the partial offload of bigger models that would not fit inside the device’s vRAM. Your RAM and disk read/write speeds have an impact on performance in this mode.

Image Quality Enhancements for AMD Ryzen AI Users

AMD Ryzen AI 300 Series CPUs have also seen significant improvements in image quality. Now completely supported are checkpoints and very efficient designs, such as Stable Diffusion Lightning, which will be the default “HQ” mode while Amuse is in EZ Mode. While offline and on the go, this upgrade significantly improves the quality of local AI picture synthesis on thin and light platforms.

Additionally, a 4x increase in native canvas size has been made to 1024×1024 pixels (~1 MP), which, when paired with AMD XDNA Super Resolution, produces an output of 2048×2048 pixels (~2 MP).Image Credit To AMD

These modifications allow Amuse 2.2 Beta’s new HQ model to produce high-quality photos with noticeably improved visual integrity and rapid adherence.

The images are much more cohesive and detailed. There is now a “boost quality” option in the HQ mode as well, which will often result in even higher image quality at the expense of longer creation times. On an AMD Ryzen AI 300 Series laptop, download Amuse and set the quality slider to “HQ” to test out this new HQ model.

Additionally, the AMD XDNA architecture that underpins Super Resolution features has been tuned to enable end-to-end upscaling at a far quicker rate than previously possible. Amuse is an extremely user-friendly AI toolkit that requires no technical expertise and automatically configures itself, especially when combined with the AI Filters and AI Design tab, which have strong AI assistants.

How to Use AMD Radeon Graphics Cards with Ryzen AI to Run Stable Diffusion 3.5 Medium, Large, and Turbo

How to install Stable Diffusion 3.5 on Ryzen AI 300 Series CPUs or AMD Radeon 7000 Series graphics cards:

Install Amuse after downloading it.

Accept the terms and conditions in their entirety.

Click the flask icon at the bottom to select “Advanced” mode. Select “Yes” when asked.

Select “Model Manager.”

In the model manager, look for “Stable Diffusion 3.5 Large Turbo.”

After selecting “Download,” watch for the download to complete.

Make sure “Text to Image” is chosen on the left after clicking on the “Image Generation” tab.

Choose the model you recently downloaded by using the “drop-down” menu: Large Turbo Stable Diffusion 3.5.

Click “Minimum Memory Mode” to continue in minimum memory mode if you are informed that the model is too big.

After writing your prompt, click “Generate.”