Category: Collaborative Unit

For this part of the course, we had to collaborate with the neighbor course, Computer Animation. I went to padlet where all the people were posting their projects idea and I saw this one called Researcher Collab. It was sci-fi so it immediately got my attention.

I liked the idea behind it and they needed VFX people for some cool effects and lighting. I wanted to try and work with them.

Here is the animatic of the project.

We created a discord channel for better communication. And we usually had a meeting every week to see how we were doing. First, we discussed the storyboard that was already created and if it needed any changes.

We also created a OneDrive folder for easy sharing files. So after we organized our self we started working.

My part on this project is to create some VFX and also set up the lighting of the scenes.

For the VFX part I decided to work with Houdini since we are learning at the moment and I could have the support of the teacher, and also I really like the program and wanted to learn more.

This is the general overview of the node editor in Houdini.

The general workflow is importing the object, scatter to points and then with this points project them on to the floor.

This points projectets are the ones that are gonna generate the simulation of the smoke/dust.

This is the first version and its only smoke, but for previz its more than fine, the idea would be to have smoke and also dust. We learn how to do that on week 6 and 7 of the Houdini class.

After that we create the density and with the pyro solver generate the smoke, touching the right attributes to make the simulation I want to do.

In the end, converting to vdb so I can Import it to Maya where we are gonna render the scene.

Here we can see a bit more of the process:

Importing the object and unpacking it.

After that we delete some of the faces that we don’t want or need to be projecting the points.

After that we scatter to create points and create a vop attribute that we are gonna use for the expanding of the points:

Here we can see how the vop attribute works. The bigger the attribute the closer are the points, and when we reduce the value we can see that the points are spreading.

• 

• 

After this we create density and a distance attribute where the farther away the ship is from the floor, the less points is gonna project.

After this is just a manner of tweaking the settings on the pyro solver to create the desired simulation. And this is the result I got:

After this, the process is to export it to Maya and then work from there, with the shaders for the render. This is a rough render of the result in Maya. The shading is not the best, but I was having issues with it that only it looks black.

Once we are finish with the work for previz, is time to render. We had some issues with the render farm at the University. All the shots that had VFX in it were crashing all the time. We reach the conclusion that the problem was because for the VFX we used Maya 2020 and its more buggy than 2018 the one that is in the computers at Uni.

And we render out the first Act without any of the VFX, here you can see the result:

After the problems we had with the renderfarm we thought that we would no be able to render the shots with the VFX, but our team work really hard to solve it and we end up with good renders to show. And with the VFX and all the sound desing, the result is this:

As you can see there are some parts that are not rendered in Maya, those are Act 02 and Act 03 that we did not have time to finish. But we knew it was a long project so we focused only on the first act to do it properly. We didn’t want to rush it and have a bad result in the end.

This practice is part of the smart vectors practice that we also use for beauty practice. In this case we had to remove the markers. This can be achieve by normal tracking each on of the markers but that would require more time and work. With this technique we only have to generate the smart vectors that will work as a track and with that we can do one patch or the necessaries patches for all the markers together.

First I denoised and created the smartvectors.

• 

• 

Then I created the patches necessaries throw the video, I created more than one because the light was changing a lot.

• 

Then I use the smartvectors to track the clip and make the patch to be in place at all time.

After that, also using the vectors created, I’m able to pass some of the motion blurs that should be created by the movement of the clip.

• 

After that, I premultiply each patch and merge it with the original video adding some grain that we lost with the denoised.

And this is the result.

Another green screen practice. This has a better green screen but also is a bit more difficult because of the fur and white contrast.

First I denoised the clip as always and then decided to use IBK techniques to see how it work.

This is the alpha I got.

Transparency and luminosity work for integration with the clip/image I wanted to add to the background.

And here we can see it works just fine, looks promising for now.

But when I added the background I found out that the alpha had some issues, some parts were not completely white. That is because I used only one treatment for the alpha when the left part of the green screen has some shadows that made the alpha have some transparency issues.

Here we can see the issue.

When that happen I had to combine different alphas from different parts of the body. I made an alpha for the right part of the body where the green screen is more bright and an alpha matte for the left part because it was darker.

I also did a separate alpha for the ears because it was a harder edge in comparison with the fur body.

I combined the different alphas in to one, and also did some luminosity treatment.

And this is the result of the integration with the background.

Green screen can be a long process and there are different techniques that can be applied to do a good green screen extraction. Here is the original plate I choose.

I decided to use keylight node for this practice. The first keylight is to create an interior alpha matte, strong completely white so we don’t have any unwanted transparencies. Then we do a core matte, more like a general matte that we keymix with a softer matte for the hair and parts that are more delicate.

For the keylight we select the green part of the screen that we want to remove and we treat it a bit to get a better result.

This particular clip was a bit tricky since the green screen was not very good. It is important to have a good clip well lit and recorded for the green screen to work at the best.

Here we see the result of the alpha. And the different alphas combined in to one.

• 

• 

• 

After that, we can integrate it with a different background. I also treated the clip so we don’t lose any luminosity and it can be integrated better.

This is the end result:

In this practice we will learn how to do a clean plate from a video, we start with a simple one that later we will use for a collaborative project.

First we track the plate in order to create cards to use for the patches.

Here we can see that the track works and we can proceed to remove the few markers that we have on the wall and floors.

We proceed to create the patches:

The first time I tried I had some issues with the markers on the far end of the plate. This is the result, I could not remove them.

I kept trying until I found the solution, I had the card in the wrong place even tho it looks it was in the right position. After I place the card in a better position I did not have the problem anymore and as we can see in the next video, the markers are removed and we have a clean plate to work from now.

Today we are learning a bit more about projections and we will proceed to transform this still image in to a video.

This is the result of the video.

Here we can see the nodes I used. We project a frame hold and we create a Scene to be able to reproject creating a small movement with the camera.

Here we can see it in the 3D viewport.

At the end I used this video to comp a 3D object creating a track with the video I created before. And we can see the result here.

This is also a 3D tracking practice but with this one, we are collaborating with the Maya class to implement and integrate a 3D model that we came up with to after work on the compositing part of the process.

So first is what we have done already to the clip, track it, and create a clean plate of the video. All of this process you can find it in the post about CLEAN PLATES.

While we are creating the patch, we can start modeling what we want to implement in the scene. First, we looked at some references, I chose these images of a steampunk-style machine/motor.

• 

• 

This is where I start modeling it. Here you can see different stages of the process. From the first model to texture and lighting.

• 

• 

• 

• 

• 

• 

• 

After we had the model created we can export the camera and track from nuke to import it in Maya so we can render the model with the right camera and make it move along side the clip.

We also create some shadow catchers to try and match the light from the video we have. As well as some planes with a aishadowmatte material to cast shadows on the walls and floor.

As you can see I have duplicated some of the geometry to fill up more the room.

With all this set up we can proceed to render the model and after import back in to nuke and do the compositing part of the project.

Once in nuke I start to work on the roto for the clip, otherwise, the model would not look like is in the room at the end.

And then we combine both the clip with the model I created for the final result.

This is how it looks at the end.

3D tracking with nuke is a very straightforward process that is used for a lot of techniques like 3D projection, create clean patches of a plate after use it to integrate other models/textures.

Here you can see the original plate that I choose for this practice. I like it because it was an alley with some movement on the back that later can be removed or something. Also, it has a post that passes through right in front of the camera and it would be a good practice to remove it. At least the one thing I’m gonna need to do is rotoscope it out to be able to create patches behind it.

After we denoised the plate and made the right adjustments like some sharpening or grading to make it easier to track, we can proceed to do the track.

As you can see I do some treating before the track, and roto out some areas that could cause some problems.

This is the part the I did the roto were the tracker wont affect.

Then is just a manner of selecting some points and create cards or 3d objects to see that the track is working and also to later work on that to replace it with a patch. Here you can see how the track work. Later on it will be improved and you will see the result at the end.

• 

• 

• 

As you can see in the previous images, we have the roto of the post that I’m putting back on top with a merge node so it passes in front of anything I create behind him. And I also show you the solve of the track, where the purple points selected are the ones that I’m setting as ground points so it knows where the ground is, this will help places patches or anything.

Here I create some patches on the background removing the people to have a clean plate without anyone on it.

• 

• 

The image on the right are the patches to remove the people and the ones on the left are what’s the next step that I’m working on it, remove the close up post that is passing in front of everything with different projections. Here is the end result that I have now.

Smarts vectors are a way of tracking the image that in some cases makes the job a lot easier if it works properly. We are going to use it for a few practice to learn it, we are going to make a person look younger and also remove some tracker markers from someone.

The first practice is going to be beauty work. I will make a person look younger using smart vectors to track the video and after that, I’m going to create a clean plate that with smart vectors is going to move with the video. Here is the node editor:

We first reformat the plate since it was in 4K and for this exercise, I will work in 1080 then I denoised the plate, and after we create the vectors. This is the first time using this feature so here is how it looks. We can see the vectors on the viewport and the attributes that we can change. I left it to default for now.

I’m going to create the patches with a rotopaint node, using cloning and blur tool. I created a grade node, so its easier to see if the track works well.

Here we can see the difference between the original plate on the left and the modified one on the right.

• 

• 

Here you can see the parts the I modified with the rotopaint:

And here we have the original plate:

This is the end result:

UPDATED AND LONGER VERSION

At this point I modified a little bit the patch I made and also created a few more to make the video longer.

An this is the final result with different passes of the process.