Nodes Browser

ComfyDeploy: How comfyui_facetools works in ComfyUI?

What is comfyui_facetools?

These custom nodes provide a rotation aware face extraction, paste back, and various face related masking options.

How to install it in ComfyDeploy?

Head over to the machine page

  1. Click on the "Create a new machine" button
  2. Select the Edit build steps
  3. Add a new step -> Custom Node
  4. Search for comfyui_facetools and select it
  5. Close the build step dialig and then click on the "Save" button to rebuild the machine

comfyui_facetools

These custom nodes provide a rotation aware face extraction, paste back, and various face related masking options.

Comparison

Patch notes

  • 2024-06-20 - fixed issue 13
  • 2024-06-14 - Sent yolo to cuda device if available (credits: kilroythethird, thanks for the suggestion!)
  • 2024-06-14 - Dropped the requirement from insightface completely.
  • 2024-05-22 - Updated GenderFaceFilter node.
  • 2024-05-19 - Added BiSeNetMask and JonathandinuMask nodes. Careful about JonathandinuMask, it's more accurate than BiSeNet, but it takes more memory; you can get out of memory more easily with it.
  • 2024-03-10 - Added nodes to detect faces using face_yolov8m instead of insightface.

Example Workflows

  • Full inpainting workflow with two controlnets which allows to get as high as 1.0 denoise strength without messing things up. Full Inpainting

JSON workflow

Nodes

  • AlignFaces
    • inputs:
      • insightface: Use the Load InsightFacenode from ComfyUI_IPAdapter_plus
      • image
      • threshold: minimal confidence score for detection
      • min_size: minimum face size for detection
      • max_size: maximum face size for detection
    • outputs:
      • faces
  • FaceDetails
    • inputs:
      • faces
      • crop_size: size of the square cropped face image
      • crop_factor: enlarge the context around the face by this factor
      • mask_type:
        • simple_square: simple bounding box around the face
        • convex_hull: convex hull based on the face mesh obtained with MediaPipe
        • BiSeNet: occlusion aware face segmentation based on face-parsing.PyTorch
    • outputs:
      • crops: square cropped face images
      • masks: masks for each cropped face
      • warps: 2x3 warp matrix to paste the cropped face back into the original image
  • WarpFacesBack
    • input:
      • images: this is the original image
      • face
      • crop: cropped face images
      • mask: masks for the cropped faces
      • warp: 2x3 warp matrix
  • OrderedFaceFilter
    • works similarly to ImpactSEGSOrderedfilter
  • GenderFaceFilter
    • needs more work, as InsightFace gender classifier isn't very accurate.
  • MergeWarps
    • I made some changes in FaceDetailer and WarpFacesBack since I've created this one, so it probably doesn't work anymore. Needs more testing and more work.

Installation

DEPRECATED:
  • You need ComfyUI-Impact-Pack for Load InsightFace node and comfyui_controlnet_aux for MediaPipe library (which is required for convex_hull masks) and MediaPipe Face Mesh node if you want to use that controlnet.