Final Submission
Breakdown Video
Personal Showreel
Pre-production
Concept:
Logline:
A young mother tends to her children at a park, unaware that her smallest decision will soon lead to disaster.”
A mother takes a quick smoke break while her kids play nearby. She tells them to be careful not to damage the plants as they run. When one child wanders too far, she drops her cigarette, puts it out, and hurries after them. Even though she tries to be careful, the wind blows and reignites the cigarette, starting a large forest fire.
Following the themes in the brief, we chose forest fires because they are timely and everyone is feeling the effects of climate change worldwide. Especially this summer, with the drought in the UK, it reaffirmed our decision to do this project, as we watched all the grass in the green spaces dry up. This was also an issue during our reshoot because we saw that all of the grass that was lush before had turned yellow and dead.
For our moodboard, we wanted stronger contrast between the lush greenery and the destruction of the forest fire, so we looked for references that captured that feeling. We also watched The Lost Bus multiple times and drew inspiration from some of the scenes where the fire started.
Project Management:
We decided to make a Gantt chart to estimate how long things would take. What we didn’t take into consideration was that there would be personal emergencies, and unfortunately it got delayed a little bit. However, I think that staying organized really helped us get back on track.
We organized everything on Google Drive and kept a constantly updated to-do list, which really helped us stay on track and keep track of everything.
This is the entire Miro board that we used for reference. I will break down the different sections below:
We broke down the different elements we would need to put together all the shots.
This was one of the early concepts that we had, the group met up at a cafe and brainstormed the storyboard. It was later on done digitally as seen from the pictures below.
However, after that, we met with Klaus and decided to revise our story. We liked the contrast between a caring character and the destruction caused by a seemingly careless action afterward, and we thought that it would make the story more interesting.
I wrote a script to make it easier for us to break down what we needed.
We found more references for fire. Looking at the breakdowns of other films with fire VFX to see what they did. We really liked the nodes they built in Nuke for The Lost Bus, so we consulted Gonzalo about recreating one of our own. He said it was an in-house engine that was too difficult to replicate, so we decided to create simulations instead.
Location Scout:
Since we were filming a forest fire, my group wanted to film things accurately in a proper forest. We did our research and went to multiple parks in London, such as Hampstead Heath, Queen’s Woods, Highgate Wood, and Epping Forest. Finally, we decided on Virginia Waters because of this shot.
We thought that the framing of this bench was well balanced; there’s a separation between the forest and where the person would sit. So it would be more meaningful that a “harmless” action could burn down a whole forest like this. This was our reasoning for selecting the location.
Then we started to prep for the 3D assets we would need, measuring the dimensions of the bench and the distance of it from the camera so we would have a reference.
We did a test HDRI for lighting around the bench to light our 3D assets during our location scout as well.
We were also introduced to a 3D scanning software by David, and we tried scanning the space with it, plugging the final scan into Maya to see the space.
Filming
Before filming, we created a shot list, equipment list, and a call sheet. This helped us plan and organize everything we needed for the shoot and be more efficient on set. It was really a race against the sun, and we had to finish most of the shots before 10AM because everything would be overexposed by then.
This was the log of our shoot, done by Selina, our script supervisor. It was extremely helpful to have this during the CG asset creation stage because we did not remember all the information off the top of our heads.
Blender Tree Simulation
Our original intention going in was for everyone to create a tree asset and a plant asset, and then we could burn it all. So I worked on a tree getting burnt in Blender. I identified the tree species in the shot and researched it.
I also took photos of it in the actual location so I could have an accurate reference, using different angles to try to create a texture of the bark from it.
After modeling the leaf and the tree, I added texture to them in Substance Painter. I also projected the image of the leaf onto a plane and used it as the texture of the leaf.
I added smoke animation to the tree and added a shader so that I could use the smoke to be both the fire and smoke.
Houdini Simulation
Initially, I used a mannequin as a reference and scaled the bench down to a realistic size. I then created a series of bench animations from different viewpoints. Using temperature as a mask, I animated the fire spreading across the bench. I also added a ring of “heat” to transition between the normal texture and the burnt texture.
This was my final node tree for this simulation.
The red section is the “hot” zone, and I selected the points here to burn first since it would be the closest to where the cigarette is.
I then turned the bench into a volume so I could create smoke from it. This is where the problem started for me. I noticed that the smoke felt very low resolution, and when I wanted to convert it into a VDB, the pixels wouldn’t interact with the mesh properly.
I needed the smoke to collide with the bench to create a realistic simulation, and so I tried to fix this issue by making a low-poly mesh of the bench in Maya. The 3D scan had a lot of holes in it, and polyfill didn’t really fix the issue.
After fixing the mesh, I plugged it back into Houdini to try and the axiom plugin I tried using for the fire didn’t register and so I was back to square one. At that point, I’ve been testing out this simulation for a week and I didn’t want to waste additional time so I decided to go for a different approach.
I used the fire simulation that came with Houdini and plugged in my geometry. I made the fire and smoke for the bench successfully!
After this, I exported my fire simulation as a VDB so that Maggie could add it to her blender file to render it together with the bench and the tree.
Nuke Compositing
I was tasked with the last shot, where everything is on fire. The challenge with this shot was tracking and changing the lighting as the camera tilted up and down.
I started with the camera tracking, using the points to create a ground. I then rendered the camera tracking to import into Blender, where I could render the bench and tree geometry with the camera tracking, ensuring accurate placement.
While I was working on the CG assets. I started to play around with manipulating the ground. This was the initial noise I experimented with. I wanted the ground to look slightly burnt while still maintaining some of the grass details. After adding it as a card on the grass, I then fine-tuned it to make it more intentional and less patchy.
I then added some embers from real footage I found online, placing them as cards in a separate scene so that whatever I did to change the embers wouldn’t affect the noise I placed down before.
This was the plate I used to merge the CG elements onto. It also included my nodes for this.
After the bench was rendered into multiple AOV passes, I separated all the channels to grade them separately. Adding the shadow in as well. I noticed that the edges around the bench were really shiny, and I couldn’t get rid of it no matter how much I graded the gloss, so I decided to roto a section of it so that the plate would cover it, and it worked out pretty well.
Then the roto took away a section of the flames, so I used the emissive pass to add some of the detail back.
I did the same thing with the tree, combining it with the bench shot, and then moved on to adding smoke. I wanted to tone down the fire on the tree, but it caused the color to break instead of working how I wanted it to, so we will render it again after submissions!
This is what the tree looks like with the plate.
I added the cards with different types of smoke emitting from them so that there would be more depth with the smoke.
This was the final with all the CG assets added and the smoke added too.
Davinci Resolve Colour Grading
This was the reference I was looking at when colour grading, I wanted to show a warmer and inviting forest. I thought that, in addition to the greenery surrounding the frame, it would be nice to make the highlights slightly more yellow-tinted to emulate an inviting feeling.
I followed a pretty standard node tree system. I changed the color spaces from Log to DaVinci’s color space at the beginning and changed it back to Rec.709 afterward. Then I worked on making sure the video was well balanced and not overexposed.
The main challenge was matching the grades between the shots done from different cameras because their colour space was so different. The Blackmagic 4K Pocket, Lumix and DJI Osmo Pocket 3 had a very different look and I tried my best to make them look cohesive.
Critical Reflection
The forest group project was a great learning curve for me, and a great opportunity to apply everything we’ve practiced in class so far into a project. Despite working with only two other people in the group, I think we did our best to navigate all the tasks we needed to do.
My teammates were always communicative and it was easy to delegate tasks according to everyone’s strengths. Everyone was eager to work and willing to troubleshoot any issues that came up. This was very important because we had to reshoot a few times because of the location and lack of experience we had filming for VFX. Even though we planned ahead of time, location scouting beforehand and storyboarding what we wanted to capture, it was a big race against time to film everything before the sun became too bright and affected the shots. It is always a good time on set and we learn from our mistakes, if there is another opportunity to film our own footage to comp, we know what we need to capture and measure.
I think time management was definitely something we could work on in our future projects. I would like to point out that we tried our best to follow the schedule that we mapped out, however, there were a few personal emergencies that impacted this and we tried our best to catch up with time afterwards. It was quite a big learning curve using Houdini on our own for the first time so it took longer than expected to create the simulation and render it. Moving forward, I think that it would be very helpful to map everything out in the 3D space and try building with rough projections in nuke first so we know what we are working with.
My group intends to revisit this project after submissions to improve on the shots. We are going to re-render the CG assets, such as the bench simulation and the tree. We would like to fix things to make it look more realistic, mainly fixing the simulation of the flames because we know that wood definitely does not catch on fire that quickly. If given more time, it would be great to explore different settings of the fire simulation, we experimented with many different ways of creating “fire” and had to go with the best render out of the experiments so that we could start working on compositing.
We also want to add in elements such as heat distortion to make the compositing more convincing and explore making “fake normal maps” to achieve it. I would like to revisit compositing the forest floor and finding a way to make the CG elements look more cohesive with everything else, adding in embers and flames as well.
In conclusion, despite all the troubles we went through, I’m glad we were able to create this project together. I look forward to refining the project in the future!
(491 word count)