Buffer: The margin of context window between mask and actual image
Mask Expand: how many pixel expand from mask
Mask Blur: how many pixel of the mask would be blured
Output Size: The long edge of the output image.
For example: 1536 with patch_type 3:4 and patch_mode right
it would result in 768 width *1024 height image
patch_type: 3:4, 1:1, 9:16 depands on what context window type are you want.
If your mask is long rectangle, you might want to select 9:16 to match your mask
patch_mode: how would the image patch. auto would select the patch mode automatically, depends on your reference image and its mask.