This resource is intended to be used by the "ReActor for Stable Diffusion" extension, which can be installed in both A1111 SD WebUI or Forge SD WebUI and is essentially a variation of FaceSwap.
You can find the extension page here: https://github.com/Gourieff/sd-webui-reactor-sfw
With this extension you can replace a character face generated by A1111 or Forge, with your favorite one loaded in this extension (ReActor for Stable Diffusion).
The extension can use as source image, the image of a face of your choice or a face model that has been previously created by this extension and saved locally on your PC for later use, or to share with other users.
Of course there are already a lot of LORA that do exactly the same thing, but by using this extension has some advantages (and disadvantages as well).
Advantages: While a LORA is strongly tied to the checkpoint type (SD, XL, F1) and influenced by a multitude of other settings, this extension will work with any checkpoint, regardless of the settings, since the source face will be applied over the destination face only after the destination face has already been generated and finalized by Stable Diffusion.
Disadvantages: the final face not being generated from the beginning by Stable Diffusion, under the control of a LORA as a whole, may not be 100% in agreement with the image requested by the prompt, as stylistic appearance, facial expression, age, etc., but only 70-90%.
The final result depends also a lot on the model used as checkpoint, as some models already recognize the character (for example Jugeernaut, but not the only one) and this will generate from the start a face whose appearance will already be similar to the desired character. That's why I recommend to include in the prompt the name of the celebrity you want to recreate. Even if not all checkpoints have been trained to be able to recognize such names, experience has shown me that it is always more useful to do this, as somehow the shape and structure of the face generated by the checkpoint is closer to what you want to generate, thus helping the final face replacement process.
So, I personally prefer to use this extension instead of a LORA because compared to using LORAs, this approach offers the best compromise between quality, consistency and ease of use. The checkpoint choice is up to you but no matter what you use, I guarantee you will get satisfying results.
You can find a short tutorial on how to create face models for ReActor here: https://civarchive.com/articles/4264/quick-guide-to-creating-and-using-reactor-face-models
All the face models was created by myself with this extension, using the possibilities offered in the Tools section. I used public images found on the internet which I sometimes enhanced the quality (resolution, sharpness etc.) in GIMP.
In conclusion I offer these models to those who use the extension "ReActor for Stable Diffusion" and who might be interested in using such models and who don't have the time or the skill to make these models themselves.
P.S. Even though the latest versions of ReActor have been redesigned to meet the SFW standard, I suppose I don't need to remind you that replacing the face of a public person must be done with common sense, respect and with restraint, so as not to damage anyone's public image in accordance with the law.
Description
V.01 The blended face mod was trained with 40 public images with various resolutions.