Please install it directly from the github repo here: <a rel="ugc" href="https://github.com/dchatel/comfyui_facetools">dchatel/comfyui_facetools</a>, the zip file attached to this post just contains the adress of the repo.<h1 id="heading-8">comfyui_facetools</h1>These custom nodes provide a rotation aware face extraction, paste back, and various face related masking options.<a target="_blank" rel="ugc" href="https://github.com/dchatel/comfyui_facetools/blob/main/examples/comparison.jpg"><img src="https://github.com/dchatel/comfyui_facetools/raw/main/examples/comparison.jpg" alt="Comparison" /></a><h2 id="heading-4">Example Workflows</h2><ul><li>Full inpainting workflow with two controlnets which allows to get as high as 1.0 denoise strength without messing things up. <a target="_blank" rel="ugc" href="https://github.com/dchatel/comfyui_facetools/blob/main/examples/full_inpainting.png"><img src="https://github.com/dchatel/comfyui_facetools/raw/main/examples/full_inpainting.png" alt="Full Inpainting" /></a></li></ul><h2 id="heading-5">Nodes</h2><ul><li><code>AlignFaces</code><ul><li>inputs:<ul><li>insightface: Use the <code>Load InsightFacenode</code> from <a target="_blank" rel="ugc" href="https://github.com/cubiq/ComfyUI_IPAdapter_plus">ComfyUI_IPAdapter_plus</a></li><li>image</li><li>threshold: minimal confidence score for detection</li><li>min_size: minimum face size for detection</li><li>max_size: maximum face size for detection</li></ul></li><li>outputs:<ul><li>faces</li></ul></li></ul></li><li><code>FaceDetails</code><ul><li>inputs:<ul><li>faces</li><li>crop_size: size of the square cropped face image</li><li>crop_factor: enlarge the context around the face by this factor</li><li>mask_type:<ul><li>simple_square: simple bounding box around the face</li><li>convex_hull: convex hull based on the face mesh obtained with MediaPipe</li><li>BiSeNet: occlusion aware face segmentation based on <a target="_blank" rel="ugc" href="https://github.com/zllrunning/face-parsing.PyTorch">face-parsing.PyTorch</a></li></ul></li></ul></li><li>outputs:<ul><li>crops: square cropped face images</li><li>masks: masks for each cropped face</li><li>warps: 2x3 warp matrix to paste the cropped face back into the original image</li></ul></li></ul></li><li><code>WarpFacesBack</code><ul><li>input:<ul><li>images: this is the original image</li><li>face</li><li>crop: cropped face images</li><li>mask: masks for the cropped faces</li><li>warp: 2x3 warp matrix</li></ul></li></ul></li><li><code>OrderedFaceFilter</code><ul><li>works similarly to <code>ImpactSEGSOrderedfilter</code></li></ul></li><li><code>GenderFaceFilter</code><ul><li>needs more work, as InsightFace gender classifier isn't very accurate.</li></ul></li><li><code>MergeWarps</code><ul><li>I made some changes in <code>FaceDetailer</code> and <code>WarpFacesBack</code> since I've created this one, so it probably doesn't work anymore. Needs more testing and more work.</li></ul></li></ul><h2 id="heading-6">Installation</h2>You will need <a target="_blank" rel="ugc" href="https://github.com/ltdrdata/">ComfyUI-Impact-Pack</a> for <code>Load InsightFace</code> node and <a target="_blank" rel="ugc" href="https://github.com/Fannovel16/comfyui_controlnet_aux">comfyui_controlnet_aux</a> for MediaPipe library (which is required for convex_hull masks) and <code>MediaPipe Face Mesh</code> node if you want to use that controlnet. You will also need to download the <a target="_blank" rel="ugc" href="https://drive.google.com/open?id=154JgKpzCPW82qINcVieuPH3fZ2e0P812">BiSeNet model</a> and save it in <code>ComfyUI/models/bisenet</code> to use occlusion aware masks.

Files with SHA256: 847e854abf97653d45340d368a2b414002ac7f87b375727e93f71fdfd7ff31a4