All names will be clickable and link to the individuals blog
My work over the last 2 days has been rough, addressing the feedback is getting ever more challenging due to the limited amount of time and resources we have. Mazyar had a great suggestion on shot 020 to use a push in to keep some of the reaction, However as I implemented that the lack of flexibility in the footage became a problem.
In every take we have Julian picks up the box, this did not lend itself well to the jumpcut. If we had more time we could ether reshoot or roto just Julians face to keep some of the reaction. However we don't have time so we have to settle for the shorter take we showed during class.
Julian also asked me to try a series of quick punch-ins, In theory this would be more feasible while also having elements of Mazyar's suggestion. Again I tried it and found that I didn't quite have enough time to get it to look good.
With Julian working to fix some of the color issues in the lighting and texture I have to go back and remove/change a lot of the color grading I had in place. In most cases this is very nice for me as I can clean up my node tree, However there are a few spots where I have to completely redo my color work. The silver lining of course is that I can make it better.
The shot 060 popcorn is a great example Emily thought the corn was over saturated and I agree. I added my white version to the contact sheet as it was the layer I was neglecting the most and coincidently the most important one. I first matched this layer to the lightest pieces of corn, Then and only then did I go and mix in the "butter" grade. I feel like this matches a lot more.
This last week was all about finalizing the timing ( and upscaling to 1080 ). Originally I was just going to cut up the shot 020 in after effects and retime each section to shorten it while minimizing "weirdness".
The version ( as seen here https://youtu.be/ZWZiiyi2JXM ) worked ok, but was to jittery for my taste. So I spent way to long reworking my footage to create a new shot
https://youtu.be/xtAeI5E8v5U This version lacks a little bit of the charm and character but works so much better for the flow. It was relatively simple to make only consisting of 2 rotos, one for the interior of the cabinet and one for the closing door. The trickiest aspect was the shadows, I had one clean shadow from when the door closed that I re-added to the time freazed popcorn box. However the opining shadows looked completely wrong, I ended up duplicating and reversing the timing of the closing shadows for the opining.
As we approach the final I find myself spending less and less time focused on one thing. Instead I am constantly bouncing back and forth adjusting values and fixing little bugs in the node network. This is not necessarily a bad thing, as a lot of improvements happen in one sitting.
Left = "Buttered", Center = clean, Right = Mix
My first set of changes were regarding the color notes, specifically the popped corn. previously I was using one node to color correct the corn, this was serviceable but It was difficult to get the right balance between the white corn and the yellow buttery look.
I took the time to split the grade out into multiple nodes, this let me create a super saturated butter grade that then gets mixed down to a natural looking level. This layout is also easier to adjust.
Julian sent me a new version of the corn with fixed shadows and a less harsh ceiling light. The overall image looked better but I needed to go back and change my grade to bring back some of the saturation lost due to the lighting change.
The last big change I made was regarding the reflection in the bowl. The previous luminance key worked but caused some artifacts on the edge of the color grade, In addition the way I set up the tree made it almost impossible to get it any darker.
old on the left new on the right
viewport gamma set to 2
To improve my script I used an Inpaint node, This node essentially makes a matte painting for you based on the input nodes. This worked better then the luminance key alone because it inherently smooths the tones of the grade, thus removing the edge artifacts.
Todays work was simple but time consuming, I started iterating towards the final Color Grading. I started with the contact sheet from yesterdays color correction. I then overlaid a grade node, this helped me narrow in on a overall look.
I then duplicated this generic grade node to each of the individual shots and spent the rest of the day tweaking each to make them look better.
Yesterday I learned to always use the histogram, Today I learned to ignore it and trust my eyes.
Today I wanted to focus on color correction, I have been getting ahead of myself and trying to worry about color grading before I have even corrected the image.
I started by finding an image of the Pantone board we have been using. Prior to using this image I just utilized the on set color checker to set the blackpoint and whitepoint, This helped but created an over exposed image. After using the digital board to set the lift and gain the color correction felt more lifelike
Additionally, I relied more heavily on waveforms and histograms. By creating a single row contact sheet I could directly compare and adjust the correction on each clip.
These tools have helped me achieve better consistency throughout the live action elements ( although I may want to tweak the refrigerator lighting ). Tomorrow, I am going to take my corrected images and re-add the color grading. I suspect my previous issues will not reappear now that I have properly set myself up to grade.
Next week, I want to break down the CG integrated shots and correct them the same way I am doing the live action plates.
Today and Yesterday I worked on Finalizing the roto for shot 70.
The first improvement I made was to introduce an erode node. This allowed me to be more precise with my luminance key while also keeping a smooth edge on my alpha channel.
The second major improvement was figuring out subtraction masks. This allowed me to have the color not appear on the popped corn as it falls in front of the pickles. I did have a lot of issues figuring out the proper premult and masking. Prior to using subtraction masks I was using the alpha channel of the luminance key, however adding in the new masks caused the alpha to be entirely based on my masks. After a bunch of fiddling with different premult configurations I figured out that I could use the lumanance key as a mask, This paired with my erode node created an alpha that I am happy with.
Today I was able to take a break from VFX work and have some fun with sound. Many other teams had to find a sound person, however I am minoring in sound, so my team was set from the start. This allows us to work on sound before a picture lock as I can modify the sound in tandem with the picture.
Sound is broken down into 4 ( technically 2 ) Steams, Backgrounds, Music, and Sound Effects and Foley. I always do backgrounds first followed by music then finally Sound Effects and Foley.
I do backgrounds first as they are fast and cover the entire video. After listing trough and collecting a bunch of sounds from the Soundly library, I mixed 3 stock BGS into our final Background.
Following the BGS, I used the song "Stitch In Time" from "Tuned Percussion Portraits" as I felt the music built in the way we wanted.
Following the music, I went back and slowly added in the sounds I collected, making sure they matched the onscreen movements. There are still a few issues with timeing and the overall mix is still yet to be done, but for this week I am happy with where the sound is at. I want to wait until closer to picture lock to finalize the sound.
Yesterday we had our last film day, this was very short lasting only an hour. We focused on the biggest problem shots, 080 and 030.
My goal today was to get the new takes added into the video as well as work on matching the overall colors in the film.
I picked shot 080 to help develop my color grading, as an establishing shot ( and the first shot of the ad ) it set the look for the rest of the film. I used a roto and a luminance key to break the image into three parts; the background the board and the text. I knew I would want control of each element especially the text, that needs to be easy to read.
I then used the same method from Friday, creating new script versions for each different color grade. The chart on the left shows the overall color grading, we ended up liking v1 the most, however Julian wanted to see a version with the shadows tinted blue. So I created the chart on the left increasing the blue of the gain more each time. Julian's suggestion stuck and we decided to go with the V2 of the shadows.
Today I spent some time creating contact sheets for both myself and for Jingxing.
my process has been relatively simple, I make edits to the original nuke script making sure to version up after every iteration. Then I export each version out and import them into a new script where I add the text and combine them into a contact sheet. Then after the team has picked the one they like I go to that version of the script and update it to be the newest, allowing me to eazisly find it while preserving the other options.
On my end I am using these sheets to help develop the overall color grading as well as the color for the pickles in shot 070. Originally I liked V1 however after seeing the side by side I began to like the V2 better. It struck a good balance, having a nice saturation of color while not looking "Radioactive"
Today was all about going back and addressing the detail notes the mentors gave as well as experimenting with transitions.
The first thing I wanted to address was the background of shot 60. last week we were using a static image of shot 70, when we should have been matching the interior of the shot 50 microwave.
To remedy this I asked Julian to send me images of the interior of the microwave, he ( as always ) was quick to send back what I needed. used a series of cards to emulate the look of the microwave.
The door did not have an alpha channel and thusly I was not able to add the student into the background. This and color discrepancies due to lack of reflections means that I will have to return to this comp and make some improvements. But for now I am moving on to scene transitions.
Currently scene 50 ( the microwave ) and scene 60 ( kernel popping ) are connected through a match cut. This may be our final solution however Julian suggested using a pickle chip to hide a wipe transition.
I wanted to experiment with this transition, and it was simple enough to make just using a square roto and a merge node. Although I did have to combine my scene 50 and scene 60 trees into one, and retime them so the videos did not play at the same time.
For now this transition has promise, the rest of the team is working on matching the animation of the two shots. Once that is done I believe this may be more appealing then the match cut.
To end of my day I needed to fix the drabness of the pickles, The color grade itself was simple. I increased the greens and reds while taking out a lot of blue to create a great color. I then decreased the gamma to give it a darker, richer look.
The biggest issue was targeting the color to just the pickles.
A luminance key helped grab a dirty mat that removed most of the bowl and background, however did very little to distinguish the pickle from the popcorn. For that I needed to revisit my good old friend roto. luckily the image is ( mostly ) static meaning this only took a fraction of the time my previous roto work has taken. I do want to revisit this and clean up the small amount of movement.
One of the notes from my team was regarding the highlight on the bowl. The CG glass tray is more frosted then the live one and the bright highlight did not match. To target just the highlights I needed to adjust the Ranges on my color correct node. For visualization I placed a different primary color in each offset ( highlights = red, midtones = green, shadows = blue ). this allowed me to dial in the exact range I wanted to correct.
This method lowered all highlights ( which I liked ) however, I wanted to lower just the bowl even more.
To do this I started with a simple roto for a "dirty" mask, Then used a luminance key to select just the highlight.
From there it was just as simple as tweaking settings on a color grade node ( and fussing about with pre-multiply ). Also using min instead of over in the mix helped make sure that my color grade only served to darken the highlight.
Now onto the topic of roto ( note I use a high Y value for roto, so I can see better, these images are not accurate to final color )
My roto starts with a cutout of the microwave, This is helpful as it prevents my later work from happening until the bowl would be inside the microwave.
From there I create basic outlines for each of the key elements ( the bowl, the corn, and the pickle ). I use transforms to create imperfect but fast animation, then comes the "fun" part. Going back and tweaking the roto on every frame that needs it.
This morning we were able to switch out the faulty card reader and get a working one. Yay!!! however, we only had 5 hrs to turn around a comped version, Booooo. Julian was the first one in the line, he had to take the live action shots for the reflection on the microwave. This gave me a bit of time to work on color grading.
previously I have been color grading in Premiere, That is good for overall color grading but is not as streamlined for color correcting each individual shot.
(left is pre color grading, right is post)
The first thing I did was take a grade node and set the black and white points to remove the tint. However we wanted the ad to have a consistent orange hue.
To keep this hue constant throughout the entire ad, I color corrected the final shot and reversed the grade node.
by removing the hue and then reading it, I have started to standardize the look. It is not yet perfect and will be something I iterate on over the next few weeks.
After Julian sent me the 3d animation I spent time remaking my node tree. Most of my time was spent setting up the roto. The biggest advancement with this version was condensing the tree. Previously I have had to use two separate trees and combine the renders in after effects. However, this time around I keyframed a merge node to create this transition within nuke. Going forward I want to work with julian to make this transition seamless.
Our filming went extremely well, I was able to remove the back of the microwave and it worked better then I expected.
We used a cardboard box to help remove ambient light and used one of our led's to emulate the light that would have been reflecting off the back
Although the filming went well, I have been unable to do anything else this weekend. The card reader in our kit is faulty and crashes computers. Additionally, Other teams reported to us that it corrupted some of there footage. When the cage opens on Monday we plan to swap it out. That should give me some time to compile and composite a video, however it will be a tight deadline.
On the positive side, this has allowed me to get a lot of work done in other classes. So as soon as I get the footage it has my full attention.
Work has slowed for me until we get our new plates, thus my focus is on that. Today I worked on removing the back of the microwave.
I had hoped that the back panel would be separate to the main body, allowing me to remove it without cutting anything. However, I was mistaken. My Dremel kit does not have any mettle cutting wheels, and I plan on going to Home Depo to get some tomorrow.
Our team has been able to work around hurricane Helene, I myself have not lost power or water. As such work has been relatively normal.
Over the last few days, I have been aiding Julian and Jingxing iterate their shots by creating fast comps. With our final plates not being shot until next weekend our focus is on pipeline and look development; this has lead to many iterations and re-renders. On my end it has lead to a balance between quality and speed, I want to send comps that look good but my teammates are waiting on me. My target is to return a comp within thirty minutes of receiving the images.
Julian's work on shot 50 has been amazing, over the span of two days he has sent me five versions. In order to stay within my 30 min I have broken this shot into two node trees, one for the roto and one for the background face tracking. Writing out these two trees and crossfading them in after effects allows for a quick visualization. The crossfade is noticeable but not jarring, and serves our needs perfectly.
Due to the weather I was unable to work with Jingxing as much as Julian. However, we still were able to get one version of the steam comped into the final scene. Although the node tree was easier for this shot, figuring out the roto was difficult. I originally rotoed around the popcorn, which made the steam feel like it was emanating from behind the bowl. I eventually realized that a large area of feathering helped sell the steam. I want to search for a more elegant solution for the final, but it worked for visualization.
Today I went to purchase the microwave from Walmart. I found the model that we were planning to purchase, however It was extremely small. After a brief discussion with my group members we opted to get a larger one and split the additional cost. The larger size will help get more accurate lighting and make integration much easer.
My role this week is to take what we learned during last weeks test filming and figure out how we will get our "final" plates ( we expect to have to reshoot our final plates multiple times ). To start I am wanting to acquire an actual microwave, one of the biggest concerns the mentors had was regarding the effectiveness of our cardboard stand-in.
I have found one from Walmart that is within our budget and am planning to pick it up this weekend.
Additionally, I found a generic popcorn box that we can show in the beginning to help narrative clarity. The only isue with this box is the kroger branding which will have to be worked around ( or edited out ).
On the more technical side I have gotten nuke set up on my home machine and started to practice rotoscoping. My goal going into the weekend is to figure out how I want to approach each shot
For the first shoot we wanted to focus on getting solid base plates to work with while also experimenting to find the exact look we wanted. For this we shot many of our scenes with multiple lenses and performances. During filming we realized that a few of our shots needed to either be split into two or merged, most notably our final product shot ( split ) and the 2 cabinet shots ( merged ).
BTS images curtesy of Jingxing Liu
After filming it was my responsibility to upload the footage to a shared Dropbox and create the first "sample" edits ( YouTube link to the version playlist ). Between each version the team gave notes on what to change. The biggest note was to remove the eyeball shot, It slowed the pacing down too much and did not give the intended effect.
The edit came together smoothly; however this is one of the first times I have used Premiere and I have had some learning pains. The edits were very dark, part of that came from the footage ( I did not have the aperture quite right ) but most of the issue was color spaces. After watching a few tutorials I partially fixed the issue but I need to go back and add some more color correction for Tuesdays class.
Today Zach dropped of the "Fancy Kit" our group will be using to film on Sunday. Additionally, I have created a Google Sheet that my group has access to. I am using this Sheet to break down every shot according to our pitch deck. I will keep a record of what is to be done (TBD) on each shot as well as the notes from the professors and mentors.
This Saturday we plan on shooting our test plates, my job for the week is to prepare the shot list and store the equipment for the aforementioned shoot. Today after class Storm and I went to pick up the rented equipment. For now I feel like I have the least amount of workload, and am trying to deal with as much of the "paperwork" as possible ( ex. creating and submitting the weekly videos ).
During the breakout portion of the first class, our team decided on popcorn. Two references were instrumental to this decision; Nile reds video "Popping corn on the cob in the microwave" that Julian found
and the Zack King "Card Freeze" video that I brought up.
The combination of these videos were the foundation of our hero shot, where a full ear of corn pops in the microwave and the individual pieces freeze mid air.
Jingxing found a prank video called "Popcorn Pranks!" that I fell in love with; and gave us a good payoff to the buildup. After the popping in the microwave escalates we cut the the exterior as the door opens and a wave of popcorn floods out.
At some point we added a pickle? I think that was Storms idea, I went along with it because I thought it was funny. :)
I think our final sequence flows quite well but we wont fully know until we start working.
Storyboard by Jingxing Liu
As the composter my main job during pre-vis was recording live reference. Recently I got a new phone that records 1/8th slow-mo with shockingly good quality. The team did arrange to get a better camera from the cage for shooting plates and higher quality reference next weekend, however what I shot this week was high enough quality for previz.
https://www.youtube.com/playlist?list=PLqrvLzDgxDMASKEN_p0Et5B860U6Q12w7
( It appears uploading it caused a lot of compression that is not visible on original I'm working to solve that issue )
Fun side note, to avoid the glass fogging up I rubbed a small layer of dish soap on the underside of the lid.
Additionally I worked to grab some textures for Julian Schenker, our modeler. We decided to use Skinny Pop as the brand of the popcorn however Julian was unable to find an adequate scan of the packaging. So to help I ran to the store and bought some.
The original photos I took had highlights that I was unhappy with so I took the images into photoshop and removed them.
Julian then added the images as a texture, and for previs they turned out well
Going forward I want to find a way to photocopy the packaging to lessen the wrinkles.