This video demonstrates the most basic method of automating facial enhancement using FaceDetailer. When generating a full-body image at a low resolution, you can observe that the face gets distorted due to attempting to create the face within a narrow range. FaceDetailer utilizes the detailer_pipe with added detection capabilities in the basic_pipe. Let's try the transformation using the 'BasicPipe -> DetailerPipe' nodes.
UltralyticsDetectorProvider offers a Detector that enables automatic detection of faces, hand, person, and more. Depending on the model type, there are cases where only BBOX_DETECTOR, which detects rectangles, is available, and models with the '-seg' postfix also provide SEGM_DETECTOR, which can detect in silhouette form. When creating a DetailerPipe, you can add prompts to substitute positive prompts. The prompts included in DetailerPipe provide wildcard functionality and LoRA load capability as wildcard prompts. Once DetailerPipe is created, add the 'FaceDetailer (pipe)' node and connect it to the node. And once you connect the original image for facial enhancement, the preparation is complete. FaceDetailer utilizes the Detector provided by DetailerPipe to automatically detect regions like face, hands, etc., and outputs enhanced images. 'cropped_enhanced_alpha' output makes it easy to visualize the images that have actually been enhanced and will be composited. The 'image' output is the result of enhancing and re-creating the detected regions. When you run the workflow, you'll see that only the facial regions are automatically recognized and regenerated, focusing on enhancing the facial area. Upon closer inspection of the generated image, the chin area appears somewhat awkward. The shape of the chin has been altered due to a relatively high level of denoising, as the original chin was beyond the scope of improvement.
but that's because the original chin was outside the scope of improvement. To address this, you would need to increase dilation to expand the area to be improved (mask), allowing it to cover and rectify the chin area. Now you can see the improvement extended to the chin, right?
