Transcript of Face Swap with Wan 2.2 Animate - Object Selection / Mask

So, there were comments about not  being able to select specific body   parts when working with “Wan” animate workflow.  Let’s look at it quickly, first I will bypass the  below k-sampler to stop the generation process. And, click new canvas to remove all the circles. The comment I remember was about selecting the   face, so I assume you must have put the green  circle on the face and the red circle on the body.  After running the workflow  the entire body is selected. In the point editor node, these negative  coords are the values for the red circles.  So, connect the negative coords with  the segmentation’s negative input,   and run the process. Now the face has been   selected, leaving the body as it is. I will try running the workflow and process   the video, after 115 seconds, I got the result. As you can see the face from the above image has   been added to the character in the video. You can also change the   background in this workflow. Here, the face area is masked and   the unmasked area around the face is the original  video which remains as it is in the final result. This is where the unmasked area comes  from, I will break the connection. In this example I have a background  for this scene as a video,   and I will use the load video node  to bring the video in the workflow.  I will match the resolution, the frames  used from this original video are 81, so,   I will need 81 frames out of this 901 frames. I will skip the 820 frames which   leaves 81 from the new video. So, that should be it, now the mask should   be kept on the new video I have uploaded. Let’s check what it gives. This will give me a flying head in between the  flowers, I will rather have the entire body   by removing the negative connection. This should generate a proper video.  After 96 seconds I got the result, and now  the movement and expression are from the   original video, the character is from a different  image, and the background is from somewhere else.  This point editor has problems because  the node is still in progress. It may   be irritating sometimes but it should be  useful when the development gets complete.  I will take this section, and  open a different workflow.  I took a few images from Civit AI, And  this workflow is the same I copied,   I have added some preview nodes to  understand how the selection works.  So, I want to select both  the wheels and not the car,   which works after having both the  positive and negative connection. Now if I want to select something  else, like these windows  And run the workflow, the  entire car has been selected.  Now we may think about filling this red circle  all over the car, but that wont work properly.  I think this “individual  object” option works better. The objects are selected, it should be  visible properly if I bypass the blocks,   and also this grow mask. Now it is visible.  We should use this “grow mask” node to avoid  the small spots visible in the black area.  Now this looks clean. The other way is by removing both   the positive and negative connections.  And using the mask feature for images. This works in some cases,  but it also fails sometimes  So it did not work, but this  worked for me in a few cases,   especially when the object is  plain, and two-dimensional. So that is it, happy creating, I  hope you have an excellent day.