top of page
Lit_0001.jpg

Full Video

Breakdown

Technical Guide

     I started by using the flip object shelf tool with a sphere as the emitter. I animated the emitter to move side to side a little bit so the stream wasn’t steady the whole time. I connected a volume source node to the last input on the flip solver. I then changed the particles from sprites to particles. Then I added viscosity (flip solver - volume motion - viscosity - enable viscosity and viscosity by attribute. Flip object - physical - change amount of viscosity). Number needs to change drastically to see results. I modeled my candle and used a boolean to create a mold for it. The mold would act as the collision object for the fluid. Once the fluid fills the mold, the mold would break apart in 2 separate pieces. To make sure that was a dramatic separation, I made the candle a collision object. They jump apart immediately due to proximity and collision padding. A wind force was also used because one of them wouldn’t fall after jumping apart. 

     To create a candle flame with Sparse Pyro is different from Legacy Pyro. Legacy has a shelf tool that you can use and even a simple way to add the flame to the billowy smoke shelf tool. However in Sparse Pyro, you have to add a “burn” source in the source volumes and set its target field to “flame.” Set the operation to maximum. You must also dive into the billowy smoke then pyrosource and add temperature and burn as attributes. Then go into the volumerasterizeattributes node and add them as attributes there. To get my flame shorter, I went into the pyro solver and turned the temperature diffusion and cooling rate up. I also turned the viscosity on. You don’t need a lot in this one. I turned off my wind and tuned down the amplitude (attribadjustvector_velocity) to get less movement in the flame. I then duplicated it so I had one for each candle. 

     I rendered all of these as separate layers, turning things on and hiding them when necessary to get the lighting. While I did have to put a lot of different layers on the farm, I thought this was beneficial because you can see which samples are actually helping the different objects. Some of them didn’t need a lot of samples while others needed plenty. I also cached all the simulations before putting them on the farm. For pyro and fluid, I had to make sure that I decreased the right things like particle separation and voxel size closer to the end. It helps the overall look of the simulation, but it takes much longer to load/render. 

     I then took all of my layers into Nuke and pieced them together using merge and roto nodes to mask out the areas I didn’t need. I used a grade node to give me transparency at the bottom of the candle flame. The grade mask was connected to a roto that was being blurred and had its own mask connected to a ramp (allowed me to get a gradient). I then merged in another roto to push this further. Finally a color correct with a roto mask to apply a blue hue to the bottom of the flame. I also cropped the render and changed the flame start times so they weren’t moving in unison.

Screen Shot 2022-03-14 at 6.44.24 AM.png
Screen Shot 2022-03-14 at 6.44.37 AM.png

Problems Encountered 

     I encountered many problems with the project, starting with my modeling skills. I really liked my candle reference, but wasn’t able to get the smooth curves it had. I ended up using the curve tool for the overall shape and revolving it. Then for the indent, I used the curve tool and connected it so I got a plane. Then I extruded that, beveled the edges, and used a boolean to carve that shape into the bigger piece. There are still some edge issues with this method, but it got me the best result. 

    I had issues with my fluid and trying to get it to look right and stop at the right time without overfilling. To get a thinner beginning and end, I animated the size of the emitter so that when the flow started and ended, the sphere was smaller. Changing this impacts the time it takes for the mold to fill up so there was a lot of back and forth to try and get the right result. I ended up not liking how long it took for the mold to fill so I clipped the bottom of the candle so that the mold wouldn’t have as much depth. This helped with time but caused a sinking issue. Once the pouring would stop, the particles would settle and sink. Decreasing particle separation prevented this (went from .05 to .02). After caching the simulation, I used a retime node to slow the stream down. 

Pyro ended up causing me more problems than expected. Simply getting a flame seemed harder than it needed to be. Once I did get it and got the behavior I wanted, I lowered the voxel size and increased substeps so that the render would look better. It ended up looking very bad and I have yet to figure out why. It gave me cleaner edges, but was giving me weird shapes as the flames. So I had to settle with a render with fuzzier edges but a better overall shape. 

Screen Shot 2022-03-29 at 11.31.28 PM.png

©2026 by Deja Smith

bottom of page