This blog post will serve as a process blog for multiple projects that I will be developing and completing in accordance to SCAD's VFX 420 Class: Technical Direction for Compositing, taught by Professor Bridget Gaynor. The post will be broken into 3 sections for 3 projects, all having their own updates that will be posted in chronological order, with the older posts and projects at the bottom, and newer at the top. Projects are separated with a double line divide, whilst updates are separated with a single line divide. Quick Click Links to each Project Blog Section: Project 1
Project 2a
Project 2b
Project 3
Project 1: Lighting CG Objects through Compositing
For this project, we will learn how to control CG lighting using shadow and light passes with compositing techniques, as well as rendering 3D elements in layer passes. Here are the project requirements:
Modeling Requirements:
You will need a grey sphere that will be animated. It should look like the sphere that will be photographed with background reference.
You will also need to model an additional round object that has some detail.
Object will need to interact with their environment by casting shadows and reflecting, therefore you will need to model parts of the environment where the object will interact.
Object must take up at least 25% of the frame.
Project Requirements:
You will select a background element, HDR and reference photography from the Project1_Resource_Library.
You will match the camera in 3D, as well as the position and the value of key light and shadow
Composite should show accurate integration into the live action plate of both the grey ball and the other round object.
Object should move through the scene showing an accurate change in lighting and shadow.
Must show how elements are being used in a buildup.
All elements must be free of rendering artifacts such as sampling noise, faceting, and any other types of "noise".
I plan on shooting my own live plate and HDRI for this project.
Week 2: 1/14/25 (Testing AOVs and Comping)
Today I wanted to create a test Beauty pass with AOVs (Direct Diffuse, Indirect Diffuse, Direct Specular, Indirect Specular). I learned that the wooden cube used for my CG Camera match is actually 5 inches big, instead of the 2 inches that I entered in. However, when making the cube 5 inches, the scaling was too big to where I could not see the whole cube:
Clipping Issue
I decided to cut the scaling in half, and change the measurements to 2.5 so I could see the whole cube and correctly camera match. After that, I just realigned the Skydome to match the chrome ball image that was taken. The Ball on top is the CG, whilst the one that is mostly covered is the original.
Trying to match Skydome placement to ChromeBall Image
I noticed that the CG chrome ball was darker than the chrome ball in the image that was taken. I changed the intensity of the Skydome from 1 to 2, and achieved a matching look. Changing Exposure from 1 to 2 instead of Intensity also created the same effect.
Matching Skydome Exposure to Chrome Ball Image
I then realigned the Keylight (CG Sunlight) so that the shadow would also match the original image. The blue markings on the left pane in the image below are to show where the shadow is from the original image.
CG Shadow Matching
I then created the CG Gray Ball to begin matching the key to fill ratio of the CG Sunlight.
CG Gray Ball
I adjusted the intensity of the Keylight to match the ratio.
Key to Fill Ratio Lighting Match (Left Pane: Original Image, Right Pane: CG Gray Ball on Top of Original Image)
After I was satisfied with what I had as a test, I then animated a ball to roll through the scene. I rendered out a Beauty pass with AOVs, as well as a separate shadow sequence. I then brought everything into Nuke.
Nuke Script
I made some adjustments to the AOV layers, mostly just to mess around with them and learn how to adjust them. I also got to incorporate the shadow plate that I created earlier. Here is the final result:
Test Sequence Render
Though I am happy with this as a test, I believe I can keep working on it to make it look more believable. I do not want this as a final version.
Here are my next steps to improve:
-Fix Shadow Plate (cracks are not lining up correctly)
-Redo the Shadow renders (instead of using Edgeblur in Nuke, create the same effect in Maya)
-Change the Beauty Object and Animation
-Render higher quality
Week 2: 1/12/25 (Beginning of the Project)
I went with Sydney last week (1/8/25) to shoot our own live plates. I wanted the challenge of using a plate that hadn't been used already, and these are the images we ended up with!
Clean Plate
Chrome Ball Plate
Cube Plate
Styrofoam Ball Plate (In Place of Gray Ball)
However... we did forget to take a shadow plate image (where the area is completely blocked off from light so the entire ground surface is covered in shadow, used for integrating CG onto the cleanplate), so I had to create my own shadow plate!
Created Shadow Plate
I created the Shadow Plate using Photoshop, and this took a while as there were a lot areas that needed to be filled in. Though the Shadow Plate isn't perfect, I feel that with some adjustments in Nuke, I can make it work. The plan is to use masks from the CG shadow renders and pull from this plate, but not using full opacity. I will then either color correct or grade to blend the images together. We shall see if this works! But for now, I need to set up the scene in Maya.
Maya Screenshot of Cube Matching (Wireframe)
Close up of Cube Matching (Wireframe)
Close up of Cube Matching (Filled)
I matched it to the best of my ability, but I currently do not know the cube dimensions. Once I get the dimensions from Sydney (as she owns the cube) , I will make adjustments. I did know her focal length and camera used, which was 18 mm and the Canon Eos r7. After some research, the crop sensor for this camera was 1.6, so the camera focal length in Maya was calculated to 28.8, which is what I changed it to. (Same steps as in week 1!) I then switched out the projected image on the image plane to the chrome ball, and camera matched that. However, we also did not take a close up of the chrome ball in place. I had to crop the original wide shot of the chrome ball to use as an impromptu HDRI for the Skydome light to reflect off the CG chrome ball.
Cropped Chrome Ball Image from Wide Shot
Here is the CG Chrome Ball against the wide shot image.
CG Chrome Ball Test Render
Original Chrome Ball Wide Shot Image
I believe the main thing that is off right now is that the render is much darker than the original. The skydome is also slightly off. I will have to check my settings to see if I set up my Maya file correctly. I also wanted to set up the CG sunlight, so I created a ground plane to see the shadow. Here is my first attempt, left pane is viewing through RenderCam which is camera matched to the original plate, right pane is viewing through the CG Sunlight/directional light:
Trying to recreate the shadow on the CG ground plane
CG Cube on Original Cube Image Plate
The first image is the CG cube on the CG ground plane so you can see the shadow that I am trying to create, and the second image is the cube on the original cube plate that we photographed. Once again close, but I feel like I can make a few more adjustments to get it closer. The next steps are to fix the issues that I mentioned in this post (shadow matching, making sure Maya color space is correct, and adjusting skydome reflection / placement). I am hoping to also begin animating a CG sphere and test rendering layers as part of the project. Creating the shadow plate took a while, so hence my lack in new progress... but I hope to make up for lost time this week!
Week 1: 1/7/25 (Practice)
Though I have not shot my live plate and HDRI yet, I thought it was good to practice 3D camera matching using images that Professor Gaynor has supplied us. Here is the original image that I have chosen:
Image provided by Professor Gaynor, shot by Jung Woo Jin
Here is my CG Camera Match using a 2 x 2 x 2 CG square in Maya. The second image is Wireframe:
CG Camera Match with 2 x 2 x 2 Square
CG Camera Match with 2 x 2 x 2 Square Wireframe
To achieve this, I first looked at the original image's properties.
Image Properties
I made sure that my Render Camera in Maya had the same resolution as the image (5760 x 3240), then I looked up the crop factor for the Canon EOS 5D Mark. Luckily for me, the crop factor is 1, so I did not have to change the focal length of the camera There was no data for the 35mm focal length, so this is what I went with. Though it lined up pretty close, I ended up having to slightly tilt the actual camera on the z-axis, as tumbling was not enough alone to match the object.
Screenshot of Maya Render Camera Transform Attribute
As you can see in the screenshot, I did change the pivot point to the corner of the square, and had it placed at the origin. You can also see the slight tilt on the z-axis in my Camera's Transform Attributes.
Though I believe this is an okay start, I plan to shoot my own plate and HDRI tomorrow!
Комментарии