Multiple Face Swap Video: ReActor & NextView Extension (A1111)

In the world of creative video content, the possibilities are boundless. And now, with the advanced technology of Stable Diffusion, harnessed through the innovative ReActor and NextView extensions, you have the power to effortlessly produce a captivating multiple face-swapped video. Join us as we explore the exciting realm of seamless face transformations and discover how you can bring your imaginative visions to life with ease. Lets get started with this journey

Before diving into the world of multi-face swapped videos, it's essential to set up the necessary extensions to streamline and enhance our workflow. Below, you'll find a list of the essential extensions you need to kickstart this exciting journey effortlessly.

Requirement 1: Initial Video with multiple personas/faces

To initiate the creation of our multi-face swapped video, it's essential to have an initial video prepared. In this tutorial, we'll work with an initial video featuring two personas or faces. Subsequently, we can leverage the NextView and ReActor Extensions to execute the face swaps.

Requirement 2: NextView Extension

The NextView extension is a valuable tool for converting the initial video into an image sequence, which can be used in img2img batch processing in combination with the ReActor Extension to generate multiple face-swapped images effortlessly. In the last step, you can effortlessly convert the face-swapped image sequence back into a video using the NextView Extension. More information can be found on the NextView Github Page. Let's begin the installation process of the NextView Extension within the Stable Diffusion user interface:

• Launch your Stable Diffusion Web UI.
• Select the "Extensions" tab.
• Go to the "Install from URL" section.
• Paste this URL: https://github.com/NextDiffusion/next-view
• Click the "Install" button.
• Once the installation is complete, head to the "Installed" tab.
• Click on "Apply and restart UI."

Note: Make sure that you have ffmpeg installed on your system. If you're uncertain about how to install ffmpeg on your device, we recommend referring to this tutorial for step-by-step ffmpeg installation guidance.

Requirement 3: ReActor Extension

The ReActor extension is crucial for enabling face swaps within Stable Diffusion. By allowing the assignment of multiple faces, it becomes a breeze to generate numerous face swaps. If you haven't installed the ReActor extension yet, don't worry; we have a detailed guide on how to install and use it. You can find the tutorial below:

How to Face Swap in Stable Diffusion with ReActor Extension

Before moving forward, it's essential to verify the correct configuration of specific global settings. These settings in Stable Diffusion hold significant importance in ensuring the accuracy of names for the face-swapped image sequence. This precision is a critical element for seamless use in the NextView Extension. Additionally, the image sequence should be in PNG format for the best results. These steps are crucial for the creation of an outstanding video.

• Navigate to the "Settings" tab within Stable Diffusion.
• In the left sidebar, select "Saving images/grid."
• Confirm that the "File format for images" is configured to PNG.
• Ensure the "Images filename pattern" field is left blank.
• Make certain that the option for "Adding a number to the filename when saving" is unchecked.

Once you've adjusted the settings accordingly, choose "Apply settings," and then proceed to click on "Reload UI".

The process will begin once all the necessary prerequisites have been successfully installed, starting with the access to the NextView Extension.

• Access the NextView Extension from within Stable Diffusion.
• Go to 'Video 2 Image Sequence' on the left side of the NextView Extension interface.
• Import the initial video onto the canvas within the NextView Extension.
• Click the "Generate Image Sequence" button.

After the video is transformed into an image sequence, you'll receive a file location that holds all the extracted images. Make sure to copy this location for employing the image sequence in batch processing within the Img2Img tab. Let's now move on to the next step

Having copied the file location containing the image sequence extracted from the video using the NextView Extension, you can now navigate to the 'batch' tab within Img2Img.

Img2img batch settings (Stable Diffusion)

• Copy the file location that contains the image sequence extracted from the video using the NextView Extension.
• Go to the Img2Img tab.
• Within the Img2Img "batch" subtab, paste the file location into the "input directory" field.
• Select a fitting location for the "output directory" (where all the batch face-swapped images will be saved).

Please remember that the location you specified in the 'output directory' will be employed later for converting the images into a video using the NextView Extension.

Img2Img global settings:

• Checkpoint: Realistic Vision
• Sampling Method: Euler a
• Sampling Steps: 30
• Denoising Strength: 0 (Denoising is set to 0 to maintain the image sequence as is, with no alterations except for the face)
• Seed: -1
• Width & Height: Ensure that the video ratio matches the image sequence you provided earlier, in our case, the ratio is 9:16.

With these settings configured, we can now proceed to the next stage, where we will utilize the ReActor extension for multiple face swaps. Hopefully, you have some exciting faces in mind for this multi-face swap! Let's continue.

Scroll down and expand the 'ReActor' dropdown menu. In this section, you must "Enable" the ReActor extension and add an image to the canvas featuring two faces.

Our canvas features two stunning personalities, Natalie Dormer and Jennifer Lawrence. We'll assign face numbers to the image on the canvas in the input field labeled as "Source Image", starting from 0 and going from left to right. In our case, at index 0, we have Jennifer Lawrence (left), and at index 1, we have Natalie Dormer (right).

Subsequently, we will apply the same face number we assigned to the 'Source Image' to the 'Target Image'. For our desired result, we want Jennifer Lawrence (index 0) on the right and Natalie Dormer (index 1) on the left, so we will set the numbers as follows: "1,0" in the Target Image (result) input field. Please note that the face numbers are separated by a comma.

Next, click the 'Generate' button and patiently wait as your image sequence undergoes the multi face-swapping process using the facial images you've supplied on the ReActor extension canvas.

After the rendering process is complete, proceed by copying the file location of the output directory where the face-swapped images have been saved.

• Copy the location of the 'output directory' within the img2img batch tab.
• Access the "NextView" extension from the top navbar in Stable Diffusion.
• Within the NextView Extension, select the "Image Sequence 2 Video" tab.
• Paste the file location (Output directory) into the "Image Sequence Location" text field.
• Choose the Frames Per Second (FPS) that suits your preference.
• Click on "Generate Video" to transform the face-swapped image sequence into a video.

Please keep in mind that the resulting video will be saved in the following location: 'stable-diffusion-webui > extensions > next-view > output_videos'.

If everything proceeded as expected, you'll have successfully created a stunning multi-face swap video like below:

In summary, crafting a multi-face swapped video within Stable Diffusion is an unforgettable journey. Empowered by ReActor and NextView extensions, it becomes a truly creative adventure. These tools enable you to seamlessly transform faces, maintain the integrity of your original sequences, and produce captivating multi face swapped video content. Whether your goal is artistic expression or adding a touch of humor, multi-face swapping presents an abundance of possibilities for creating mesmerizing videos. So, embark on this exhilarating journey and unlock the endless potential of content creation.