Smart Vectors and High Frequency Separation

Final result and breakdown:

I started with the creation of the Smart Vectors, which is a straightforward process of plugging your footage in and writing it out.

Next, I used high frequency separation to get the lighting and color separated from the fine details, such as wrinkles and pores.

This is done by blurring the details out of the footage, then dividing it by the original. I used a rotopaint on the details to smooth over wrinkles while keeping the underlying skin tone. I then used multiply (the inverse operation), to put them back together. I created a roto mask of the areas which I wanted to isolate (seen in the video as highlighted in red) and a SwitchMatte node to copy that to the painted clean plate. SwitchMatte is essentially just a copy node, it is part of the Luma Pictures toolset from Github.

From there I fed the clean plate into the Vector Distort node and read back in the exported Smart Vectors. The only settings I changed was the reference frame to the frame that I painted my clean plate, and the frame distance to 32. Then it gets over'd onto the plate.

• 

• 

• 

• 

Now it's time to bring in the Nuke symbol which will serve as the tattoo. I did some color correcting to help it feel right in the plate. It felt really flat and just 'slapped onto' the actor's face.

After consulting reference I noticed it needed some pores or skin imperfections on top of it. To accomplish this I used a noise pattern to create highlights, transformed it down a pixel, and used that to create some shadows for fake bumps.

• 

• 

It's a technique I utilize a lot when making detail to give the illusion of depth.

Because I used frequency separation earlier, I already had his pores separated from the plate, I subtly multiplied this onto the tattoo to bring back some of the actual skin imperfections.

I picked a frame that I would use as the reference and transformed and warped it into place. After that I added the Vector Distort which took care of the rest.

My overall node tree: