Category: Nuke

As previously mentioned in my Maya development post due to delays in the project we were meant to be assigned this week this class would instead be used to catch up on projects and have access to Gonzalo for questions for the full four hours of the class. It would be an understatement to say how much of a relief this session was for the class and I. Nuke is a very complicated software that I struggle to find tutorials about online so having this time to address our questions to Gonzalo and get them answered was a fantastic opportunity.

With the additional time I also made the choice to restart my garage project from scratch as with the knowledge and techniques I had now I knew I could do it better and that it would be much easier to just restart it from scratch than to go through it and fix the issues present.

To approach it differently this time I decided to break everything down into tasks and iterate forward from each task. I made a clean plate and exported it, I then roto’d the clean plate which allowed me to avoid the issues I had last time with the clean plate being separate from the roto and I had. With that all done I was able then to take over some of my work from my previous file and render out this new version.

I also worked on the assigned creative piece to add more to this scene of a Devil Man.

After tracking his horns and adding fire to its tips, which I may try to use a displacement map or something like it to make the flames flicker more, I wanted to try using STMaps. While I had experimented on using it with last week’s marker clean up it didn’t really work. This time however it seemed to function a lot better seamlessly ‘tracking’ a crack and soot texture I applied to the Devil Man’s face, alongside having rotobrushed out his eye and making it ‘glow’ using the dodge tool.

This time using the smartvector and STMap was incredibly easy and had no issues at all. I think its due to the fact that lighting in the scene doesn’t change very much at all.

In class we went over, frankly, a lot of information in regards to Motion Vector and utilizing different nodes in Nuke to perform more complicated cleaning and roto work and automating it. Alongside that we were learning a lot about using Motion Vector to create STMaps and use that to feed in information to create more complex and automatic rotos to create and attach VFX. That being said while I understood using tracking elements for roto clean up when it came to the more complicated nodes and using STMaps I am frankly a bit lost. I will need to review the lecture in order to get a better understanding as it is a lot of information about complicated subject matter on Nuke.

A technical homework piece was to clean up the trackers on an actor’s face. Which is here down below

While I originally thought it would be quite easy there turned out to be some surprising difficulty as with the motion sometimes areas that I had cloned would end up patching the background in on the face of the actor. As I was so used to using rotopaint as a still I didn’t think that I needed to be wary of what part of the face I take from as she tilts her head around meaning what used to be skin becomes background foliage.

Using techniques I learned in class with multiple keyers, alongside make a hard matte, soft matte and transferring the alphas from these to a despilled version of the green screen. Though I still need to review how to make proper soft mattes as it is clear that I lost some info in the hair of the woman. The lighting as well doesn’t match up as the strong orange light in her hair doesn’t mach with the scene.

For our ongoing project I decided to try implementing a 3D object that I had previously made in the course into the scene to try my hand in 3D object implementation. One of the biggest struggles with it was adding a shadow to the scene as I had originally tried a few different ways that resulted in issues with the merging. Namely it would ‘double’ up the floor making a square piece of it, that was a card that the shadow was projected onto , to be significantly darker than the rest of the scene.

Though after examining another script that was sent to us by our instructor I was quickly able to fix that issue. I also brought in a 3Dtracker from a previous version of the file to use that to generate the cards I would project the surfaces onto which did fix some of my issues when it came to objects moving around in the scene when they should be totally still.

Again issues exist with my current roto but I still lack the time to dedicate to fixing it.

This week we focused on green screen techniques. The multiple nodes that Nuke provides for removing green screen and how they differ. With that in mind we also went over several green screen techniques to fix common issues like spill. Alongside that we learned how to average out a green screen to make it easier for the nodes to remove it cleanly. That being said I will have to go back and review these lecture notes as it was a lot of information and some techniques weren’t properly digested. I would say this is my largest and most consistent issue when it comes to learning Nuke. Given our short time in class there’s very little time to ask questions about specific issues we face in Nuke. These issues are of course personal and from a variety of potential problems that need time to be tackled and may not help most other people in class.

We had homework to test out our green screening techniques. Pictured below is my final alpha for the green screen which to me is pretty clean. There do appear to be some issues with it though it doesn’t appear to be reflected in the final comp so perhaps they aren’t an issue? One thing I was wary of is that the green box on the machine behind the mad scientist was having its green removed. To fix that I just roto’d the box and tracked it very simply and after the scene was green screened just placed it on top so as to maintain the color of the box. One big issue with this is that the de-spill hasn’t worked as intended and you can still see some green on it especially when his arm swings and you can see green around the balls.

In my garage this week I was aiming to add texture to the wall of the scene to make it more grimy and rundown.

While for the most part its successful I have a continued issue where my roto shifts and moves around so it looks strange and as a result elements in the scene like my poster and spray paint don’t remain fixed in place. While I try to fix it through importing the tracking marks from previous files and making my own again it results in even worse tracking than when I simply placed the elements myself into the scene. I also lack the time to redo the rotoscope with proper techniques and to experiment to fully grasp nuke’s 3D scene roto.

This week we focused on CGI compositing, specifically the titular machine this project has been named after. The machine was already set up in its EXR to match the movements of the scene it was placed in perfectly so matching it was not our concern. Instead our concern was compositing via grading, color matching and lighting. The lesson went over using the multiple passes, tips and techniques on color grading and a focus on understanding that where you take your grades from is important. A single shot has multiple lighting conditions.

As an example in our project the front room is brighter than the back. So if you graded the machine based on the front room then the machine would look out of place as it doesn’t match the lighting conditions of its actual setting in the scene.

To encorporate my work into the scene provided with us I exported my own clean plate and in addition to my clean plate I exported my rotowork as its own layer. By layering these I was able to make it that the machine appeared behind the wall.

That being said it taught me something quite valuable that I was missing. Breaking up my work into small chunks. I had issues previously in my roto because of the fact that as I struggled to do it all in one scene and so the roto would take the uncleaned plate and that would mean to achieve the final look I had to shift elements around in 3D space when I shouldn’t have needed to. Instead I should focus on completing scenes step by step. I should have, after cleaning the plate, exported a version to then serve as the roto plate. In future I will do my best to keep things simple and layered.

In addition to what had already been done for the scene we were tasked to add to it. As I thought it would be useful to know what the machine does to create a story around it to influence the scene I had asked about it. There was no such purpose to it so rather than come up with one I decided to create a scene around keeping it vague. So far I have added a bloodstained graffiti begging no one to turn it off and created a poster in the scene. I still need to add more but I am thinking on what further additions I can make.

This week we focused on advanced components of EXR via their multiple passes that allow us extensive lighting controls in Nuke. Alongside that how using different methods we can grade those different layers safely without affecting other portions of the grade or compromising the alpha. For homework we had to use the different passes to grade a car into a scene and make it look realistic. This was my result.

This was not too difficult as it was only a still though I think there’s certain way it can improve. Namely I’m not sure if the lighting is correct on the top of the car. I think that maybe, depending on distance, the lamp on the bus stop directly behind that car should light up the roof of the car a little but I’m not too certain.

The hardest thing for me to do this week was re-arrange my scene from last week as essentially the way I had done it, thinking it would be simpler with the merge, resulted in a scene that would only be fixed if I implemented 3D geography and matched it to the bricks and such because the cards, being 2D, did not have enough geometry for it to function. Having to re-do, re-arrange and re-think how I did the scene was difficult as its painful to have completed something and then have to essentially restart.

Graded

As can be scene there are still some issues with the roto, namely as it gets closer the quality of the image becomes blurry because of where I took the projection. It can also clearly seen to be flat as well so this would require me to, at a later point in time, go back and make adjustments to the roto. This however need only come later as there are more pressing things that require my attention.

This week for Nuke Development in class we went over more advanced Projection techniques to use in our 3D match move and had to apply them to our ongoing garage project. Getting my head wrapped around using 3D space in Nuke alongside the nodes and set ups to use these projection techniques was not easy, but after about an hour I was able to figure out, for the most part, how this set up functions and how to emulate it using all of the 3D nodes available.

Starting with marker cleanup I had certainly improved in my organization of my timeline and using backdrops to keep track of what each set of nodes accomplishes making corrections a lot easier. While figuring it out the first time was difficult once I had understood it the rest was simple

Rotoscoping while arguably a much simpler looking set up was the most time consuming, especially when it came to the finer edges of things like the lamp. Using the 3D merge node I was able to combine them to create what should function as a seamless transition between the three versions of this roto which help to fill out gaps that occur in the roto as the wall gets closer. However as can be seen there are some visual issues with this that I need to address

The visual issues are far clearer here in the rendered out video and is something that I need to address. The patches as well don’t come up as rotoscoped visually with the grade but that is because they are a separate card from the roto which if layered on top would show the markers again. However this is a purely visual issue and functionally works as a fully roto’d and functional piece.