Search
  • leighrmiranda

WEEK 6: Quality Control

Updated: 3 days ago

SUMMARY: During week 6, our group worked on improving render sampling/noise issues in the transmission of the train windows and added glow to them to make the inside of the train feel like it's lit up. For compositing, I continued to work on DOF/lens effects, made the crowd more dense and diverse (less repeating people), and adjusted colors/contrast to avoid making the shots feel overly foggy/hazy. We also found a new sound designer, Yash Patel, after our last one dropped the project, and he was able to get a first music/sound pass out for us this week!

New Visualization:







Feedback: Our group would like feedback on whether or not the train windows should be lit up. Here are some comparison images: The train shader in this one got rendered incorrectly in the second image so the first shader is the correct one.





BREAKDOWNS/PROCESS: Creating the Crowd: Last week, I didn't have a lot of time to work on the green screen and therefore wasn't able to render out the denoised sequences before the submission deadline. Because of this, Nuke was running extremely slow in projection and I had to merge and render the entire crowd onto one card to get it working and turned in on time. This week, after rendering out the denoised sequences, everything ran much smoother and I was able to put each person on their own card to begin refining the crowd and the position of each individual person/card in space. Projection Progress 4.29.21: (11 cards total)


Projection Progress 4.30.21: (33 cards total) To make the crowd feel fuller, I split my comp into 3 different scenes and duplicated each person 3 times. Using individual cards, I can move each person wherever is needed to frame the composition. I learned it's pretty tricky to place the people when the camera is so close to the stage and in the middle of the crowd because their heads get really big. After I got the placement right, I retimed each person so there's no obvious duplicating movement. After getting the basic crowd placement with these 3 groups, I realized that my organization was slowing me down and was causing confusion because I was having trouble finding where each person/card was in the node tree. It also didn't make sense to key the same footage multiple times. To fix this, I optimized the tree by only including one read node per person and color organizing/labeling each group. This made my comp much smaller and therefore easier to locate individual nodes and people in the crowd. Before Organization:


After Organization:


Completed crowd with projection and render cam visible:

Working with Color:

Here I was working on balancing my color grading on the week 5 renders to create more contrast and not cover up the black areas too much, such as with the bottom left guitar area.



After getting the new renders from Ken, I used a color correct to bring the gain of the shadows down here too. His new lighting also created more diversity in the color palette by introducing more cool tones into the shadows in the beauty pass.



Final comp for shot 1 after color adjustments, window light, motion blur, etc:


Lighting the Train Windows: This week, we decided to try to light the windows of the train to see if it created a more homey feeling based on this reference:

Here's the result for shot 3! We put the windows on a separate render layer and Amanda parented a light to the windows to move with the train. The issue we ran into is that the back windows were also showing when I merged in Nuke. We tried to fix this with an aiflat shader on the back windows in Maya. However, the aiflat was still part of the alpha so I had to resolve this by shuffling the red channel of the beauty into the alpha and luma keying it to be pure white/black as to only show the front windows. We also had another issue in shot 1 and 2 where certain windows were much brighter than others, which is likely due to the location of the parented maya light that we will troubleshoot for the upcoming week.

There are some extra train pieces accidentally added to this layer that will be fixed in the next iteration.



Thanks for reading!

56 views0 comments

Recent Posts

See All