1580158
/
batch-face-swap
mirror of https://github.com/kex0/batch-face-swap.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
batch-face-swap/README.md

2.3 KiB
Raw Blame History

Batch Face Swap for https://github.com/AUTOMATIC1111/stable-diffusion-webui

Automaticaly detects faces and replaces them.

Requirements

MediaPipe Python package To install it, just open requirements_versions.txt located in your stable-diffusion-webui folder and add mediapipe

Installation and Requirements

The face detection requires MediaPipe Python package To install it, just open requirements_versions.txt located in your stable-diffusion-webui folder and add mediapipe

To install the script just download the zipped script Here and copy the file face_swap.py into your scripts folder.

Guide

  1. Open img2img tab. (you don't have to go to the Inpaint tab, if you do, you have to set "Mask source" to "Upload mask" or generation won't work)
  2. Select Batch Face Swap script.
  3. Paste a path of the folder containing your images in the Images directory textbox.
  4. (Optional) It may sometimes fail to find a face if the face is very small in comparison to the size of the image. So, you can tell it to split the image and look at the smaller portions of the image by using the How many times to divide image slider. (don't worry it will stitch the image back together)
  5. Click Generate

If you want to adjust Denoising strength or Mask blur you have to disable the override checkboxes.

You can generate and save the masks without even engaging the stable diffusion image generation by checking the Save masks to disk checkbox and pressing Generate masks button at the very bottom.

Tip: You can check how many faces do your current settings find before you start generating with the stable diffusion by just clicking the Generate masks button at the very bottom.

Example

example

detailed closeup photo of Emma Watson, 35mm, dslr
Negative prompt: (painting:1.3), (concept art:1.2), artstation, sketch, illustration, drawing, blender, octane, 3d, render, blur, smooth, low-res, grain, cartoon, watermark, text, out of focus
Steps: 50, Sampler: Euler a, CFG scale: 7, Seed: 4052732944, Size: 512x512, Model hash: a9263745, Batch size: 8, Batch pos: 1, Denoising strength: 0.5, Mask blur: 4