How to use Faceswap in Fooocus

Did you know that there’s a tool in Fooocus that is capable of creating Faceswap? That’s right. Fooocus allows you to customize a photo using AI prompt, especially for the costume, expression, or specific scene similar to deepfake process. The great thing about Faceswap in Fooocus is that compared to another Faceswap alternative such as Roop in Automatic1111, Faceswap is way simpler and gives you great results. Even better, in DiffusionHub it is already pre-installed so everything is just few clicks away. We will show you how!

Method

• Login to your DiffusionHub account, choose Fooocus
• In Fooocus, tick “Input Image” and “Advanced”

• You will see another drop-down with other tabs, click on the Image Prompt Tab. 
• In the bottom left corner, you will see “advanced” box. Tick that. You will soon see there are other settings appear which is “Stop at” & “weights”. 
• Upload your image in PNG or jpg format. You can also upload multiple image so the AI will learn about multiple expressions. Please make sure the photo o be high quality, not blurry, and preferably portrait.  
• You can start playing around with the stop-at & weights settings. In this case, we will need the stop-at number to be quite high. We are talking about 0.95++ and weights at 0.8++. Don’t forget to also tick Faceswap. Here for example we use Blake Lively photo.

• Go to the top right at the advanced tab, you may adjust the resolution settings. For example, if it’s a portrait image, it’s best to put 768×1280.
• Write your prompt and click generate.

Here are some of the results

Blake Lively

Ryan Gosling

Pretty nifty right? Another Faceswap advantage is you get to save money instead of hiring professional photographer for a self-portrait.

Things To note

In our experiment, it is quite important to adjust the settings as you go. As noted above, Generally, the “stop at” settings need to be quite high because this settings tell you when the AI need to stop training. Since we are training the face and we want to have as similar as possible, we want the AI to train most of the face before it stop. 

As for the base model, make sure to choose SDXL as Fooocus currently only supports SDXL. Finally, if you have a specific model in mind you would like to download, DiffusionHub also makes it possible to download hundreds of models in CivitAI/huggingface. 

Another thing to note is that faceswapping is quite similar to deepfake. It is important to highlight that while we do not censor creators, DiffusionHub would like to emphasize the need for creators to respect the law and take responsibility for their creations. It is also important to note that NSFW deepfake without consent may result in law violation and distress to affected individuals.

Let’s take responsible use of AI and happy Faceswapping!

Share on Facebook

Post on X

Follow us

Save