WEEK 5: Adding Depth
SUMMARY: During week 5, our group focused on refining our renders from last week. We also shot new green screen plates and edited lighting transitions and camera/train timing. I had a lot of work in store for me this week from booking/shooting in the green screen room to lighting/rendering shot 2, compositing all shots/elements, and editing the latest visualization.
New Visualization Edit:
Lighting Shot 2:
I started the week off by fixing the lighting transition in shot 2 by animating the cone angle and rotation of a blue spotlight to move across the scene. I also added a subtle blue spotlight in front of the guitar to create a blue highlight on the body of the train. After creating the lighting, I gave the file to Gen so that he could also render the smoke and confetti elements using the lighting setup. By doing this, the light is able to illuminate all the elements and seamlessly tie into shot 3.
Overall Compositing: Shot 1 - Adding Depth:
For shot 1, based on feedback from last week, we wanted to create a heavier depth of field with bokeh using the confetti caches we had from Gen. To do this in comp, I used a separate zdefocus node for the confetti and the overall background. For the confetti, to create the circular bokeh effect we wanted, I found out that using the math "direct" type in the node worked best. In this shot, I also adjusted motion blur using vector blur and the motionvector AOV. Below, you can see the progress from the raw beauty render , beauty + atmosphere and confetti , and then final comp with motion blur, full DOF, color, and lens effects. I added a subtle lens flare by changing settings in the Nuke AutoFlare gizmo by Vincent Wauters. Originally, Ken didn't render any AOVs in the beauty and rendered confetti with the background, so I had to get creative and create a confetti alpha using the diffuse AOV with a luma key from the confetti render. Ken re-rendered after that , so this was no longer necessary, but I did need to use an old beauty render to create an alpha for the train in order to mask the confetti behind it since the train wasn't included in the confetti render and the confetti would have appeared to go over the train instead of behind it. Unedited Beauty Render from Ken:
Atmosphere + confetti:
Depth of field, motion blur, and lens effects:
Node Tree for Shot 1:
New Green Screen Plates/Crowd: After beginning on last week's green screen footage, we realized it had too much motion blur to be easily keyable or effective for this project due to poor lighting and low shutter speed when we shot it. To fix this, we decided it would be best to re shoot and I was able to get a reservation for the SCAD green screen room on Wednesday for an hour. Wednesday morning before we went, I spent a long time researching my camera settings to make sure everything would work. We also brought more outifts so that the crowd would be more varied. This time, we shot at 60 fps with AVCHD log footage to get higher quality in post. We also shot at 25 mm to match the Maya camera. The edges were much smoother this time but the only thing we didn't realize was how much noisier log footage is. Since we only had the room for an hour, we didn't have time to set up, shoot, and check footage before leaving. To fix the noise, I was able to get the help of my friend's NeatVideo software to clean it up.
Before and After Denoising:
Since we had 11 total plates, it took me a while to key all of them along with other projects, rendering, lighting, and comp. Therefore, I ran out of time to project each person on separate cards this week. Nuke was also being extremely slow with projection so I had to pre render the rough crowd and put them on a single card for this submission deadline. I ended up layering 2 cards with the same crowd mirrored just for testing depth. This week, the tracking was a lot more effective after I followed Kyle Cody's suggestion of importing the stage geo from Maya into Nuke to use for placing the card. Going forward, I want to continue refining the keys, place each person on their own card, and work on overall scale. We think that we may also need to raise the actual camera in Maya since the people take up so much of the frame at this more accurate scale. I also realized that with our strong light transition in shot 4, the people will need to be illuminated with it so I did some research on what crowds look like standing in front of bright spotlights:
It seemed to me like the rim lights on the people essentially got stronger, so I decided to recreate this effect using a lumakey and grade method in Nuke that I keyframed to match the frames where the spotlight is brightest in shot 4. Rim Light Method:
Before and After Rim Light from Lumakey:
Projected crowd in relation to the stage geo in Nuke: