A workflow that makes the image clearer.
The nodes in ComfyUI ImpactPack are more precise in adjusting and overlapping masks compared to the commonly used SDUpscale, resulting in better-looking textures. However, due to more segmentation, the time required to enhance the image is relatively longer.
The following workflow offers several options:
1. Keeping Face From Detailer: Faces are not sampled. Face sampling can be influenced by the model. Enabling this group ensures the face remains unchanged. If you want the face to be sampled, you can skip this group.
2. UpscaleUsingModel: Using a model to enhance image quality will make the image clearer but may also reduce details or introduce unnatural elements. If you notice reduced details, you can skip this group.
Additional explanations:
Resize is preset to 2400. Under this setting, square images will be divided into 4 segments, requiring 4 rounds of sampling. However, with TeaCache and TurboLora, the speed is still acceptable. If the dimensions are increased, the number of segments will also increase. If you don't mind the generation time and only want a very large image, even setting it above 8000 will supplement clear details without visible seams.
Denoise is set to 0.3, similar to FLUX's Image2Image settings. The higher the value, the greater the detail variation, while lower settings are closer to the original image.
The segment size of MakeTileSEGS can be adjusted by modifying bbox_size.