There are a lot of opticors available in Open Source Space and in this post we find and detect the head, which has been released by the Fall-I, which has been included in the compressed using the custom node. aurasr -comfyui. It is a GAN -based super resolution for real -world images, the image conditional is a different state of the Gigan paper for the conditional advanced.
There are two versions of this: The first version When JPEG files have some antiquities when promoting files but Version two leaves They found me to use version a dejpg_omnisr Comparison gives better results. The model is so mild that you can run on a low VRAM GPUS (according to my research 8GB minutes) but I only managed to test it locally on my RTX4080.
I also assume that you have a basic understanding of Comphyui and you have installed a version of it on your computer. If you are new to Comphyui, I suggest that you read this post when you start with COMFYUI.
Sequence
So that install install aurasr comfyui The node will install the desired customs nodes using a compressed manager. You can easily download the work flu and install the missing custom node using the Comphyui Manager option.
Download the model of your choice v1 Or v2 And reserve the JSON file and under it ComfyUI\models\Aura-SR Folder
Ensure to be installed once you can restart Comphyui and interpret your browser.
Now you are ready to increase your photos to 4x size. I have added a mild blurred pin because the high resolution image can be a bit faster and can fade through 1-2 radius to ensure that the icon does not look too much.
Download the workflow
You can download the workflow below, which provides the JSON file as well as PNG. You can drag one of them or open them in the comphi to start.
Aura-S. ROPSkil (48 downloads)
Once you have a workflow, you can upload your photos to the image node and run the workflow so that it starts with it. I was able to prolong the images up to 8-64 megapixels in a few minutes. You can always create big pictures because the model can handle it.
Here are some sample images that I was able to prepare using this workflow. Two of these real images are advanced and two other generative AI images are advanced. Can you tell who is?
If you have any questions about this workflow and head, tell me. It is a huge model that maintains a lot of details when promoting icon, it does not affect any Hallicanic and that is why it is good in reality.
If you want to support our site please consider buying us FiFor, for, for,. Catch a product Or Subscribe. Fast GPU is required, access the fastest GPUS for less than $ 1 per hour runpod.io