Nodes/ComfyUI-FatMex-Nodes/Fat Mex Face Swap
ComfyUI Node

Fat Mex Face Swap

Face swap with two modes: • Quick – InsightFace inswapper paste (fast, no extra model needed) • Quality – YOLOv8 detect → inpaint face with editing model + 4x upscale Connect MODEL/CLIP/VAE from a Qwen Image Edit loader for quality mode.

By FatMex·Created 6 months ago·Updated 6 months ago· 3
Fat Mex Face Swap
  • target_image
  • face_reference
  • model
  • clip
  • vae
  • swapped_image
face_index0
blend_strength1.00
methodauto
denoise0.60
feather13
steps4
upscale_facetrue
CategoryFat Mex

Inputs (12)

NameTypeDefaultDescription
target_imageIMAGEThe generated image to swap a face onto.
face_referenceIMAGEClear photo of the face you want to use.
modeloptMODELFor quality mode: connect MODEL from a Qwen Image Edit loader.
clipoptCLIPFor quality mode: connect CLIP from the same loader.
vaeoptVAEFor quality mode: connect VAE from the same loader.
face_indexoptINT00–10Which face to swap (0 = largest).
blend_strengthoptFLOAT1.000–1Quick mode: blend strength. Quality mode: ignored.
methodoptCOMBOautoauto = quality if MODEL connected, else quick.
denoiseoptFLOAT0.600.1–1Quality mode: how much to regenerate the face (0.6 = natural).
featheroptINT130–50Quality mode: mask edge feathering in pixels.
stepsoptINT41–30Quality mode: sampling steps for face inpaint.
upscale_faceoptBOOLEANtrueQuality mode: upscale face region with 4xFaceUpDAT.

Outputs (1)

NameTypeDescription
swapped_imageIMAGE