Face Toolkit
A set of tools for working with faces in video. Face Swap transfers a person’s appearance from a single photo onto the face in the clip while keeping its expressions, head position and scene lighting, useful for stand-ins, reshoots without the actor and anonymisation. Live Portrait transfers head movements and facial expressions from another video or photo, and without a source you can set them by hand with keyframes like smile, blink, head rotation. The tools change only the face and head, so hair, body shape, hands and body movement stay from the source. Strong turns into profile, a face covered by hands or objects, and fast camera motion produce artifacts. Live Portrait does not sync lips.
Requirements
Linux, Windows, macOS. 6 GB VRAM, GPU or CPU. Python 3.11. Models: ~1.4 GB download.
Disclaimer
The plugin code is developed by Wunjo. The models are made by third-party authors and come under their own licenses. Results depend on the source material and are not guaranteed. You are responsible for the content you create and for the consent of anyone whose face or voice appears in it.
How to use
Right-click on the video clip that contains at least one human face. In the menu that appears, select “Artificial Intelligence” > “Enable Face Detection”. Rectangular overlays will appear around the detected faces in the “Project Monitor”. These overlays are only used for face detection and will not appear in the final render.
Left-click on the face rectangle and select “Face Toolkit” from the menu that appears. Then click on the video clip in the timeline. The “Face Region”, “Face Swap” and “Live Portrait” effects will appear in the “Effects / Composition Stack” panel on the right side of the screen. If necessary, you can adjust the position of the “Face Region” effect on the timeline to refine the face detection results.
In the “Face Swap” effect settings, locate the “Face source” section and select the image file containing the face you wish to use for the swap. If the desired face has not yet been registered, click the edit icon to upload a new file and register the face. In the window that appears, select and upload the required file, then click “Register face”. You can also view previously uploaded faces in this same window.
To use the “Live Portrait” effect—specifically if you want to replicate head movements or facial expressions from a video or photo—locate the “Expression source” section and select the video or image file containing the person whose facial and head movements you wish to copy. If the desired face has not yet been registered, click the edit icon to upload a new file and register the character. In the window that appears, select and upload the required file, then click “Expression register”. You can also view previously uploaded characters in this same window.
You can also add emotions and head movements without a source video. Simply select keyframes on the timeline and adjust the settings for smiling, blinking, head rotation, and so on, for each point.
Once you have finished configuring the “Face Swap” or “Live Portrait” settings, click “Generate into the bin” within the selected effect and wait for the generation process to complete. Upon completion, the generated file will appear on the timeline.
How to install
Download the .wmplugin file. In Wunjo Make, open “Settings” and then “Plugins”, and click “Add plugin…”. Select the downloaded .wmplugin file. The plugin will be copied into the application.
The plugin’s Python environment and model weights are downloaded automatically the first time you run it, so keep your internet connection active until the process is complete. You can also download the required Python libraries and models manually from the plugin settings. An internet connection is required for the initial download, but it is not needed for subsequent use.