top of page
  • zarahnajmi

Avatar VFX Shot Recreation Week 12: 3D Modeling and Pyro Adjustments

Ho boy, it's getting close now.


This week, I want to focus largely on my 3D model, as well as getting in the final pyro trails and debris trails.


Adjusting the Samson's Animation for a better Pyro Sim

So, remember how last week I had to add some wind to my main pyro to force it more towards the left to provide more coverage? Another thing I felt was needed was to add more horizontal translation to the Samson in it's animation. See, in the reference, when the aircraft is struck my the missile, it is quickly shoved forwards towards screen right. I feel like I needed a little more of that shove in my animation. So, I went back and tweaked the graph editor to make these changes (adjusting the graphs is easier for me than repeating all the keyframing). I also realized I needed to bring the camera closer to the Samson, which should also help a bit with coverage. The majority of these animation adjustments were made in the second half of the animation, post-missile impact. The change looks subtle, but the difference in the helicopter's positions-by-frame and point velocity should make a difference in the pyro sim.


Finishing the 3D Model

Cockpit & Windows

The biggest thing that needs to get done on this model is the cockpit , which I’ll admit has been intimidating me for a while. But, I have my base mesh ready to use the make live tool on. I’ve gone ahead and created a rough blockout for where the windows will be cut out of the cockpit, as you can see here.



At this point I also added a bit of extra detail to the roof of the aircraft in order to better guide the window placements.


Now, it’s just a matter of going through with the quad draw tool to draw out the new mesh flow. If you don’t recall, I’ve already created this guide for myself:



Getting into quad drawing my topology and this thing is fighting me. Retopology is basically a huge puzzle, trying to achieve complex shapes while keeping topology clean enough so that it subdivides properly. Although, it’s not too bad when you get into the flow of it.


I started off by simply outlining each window with loops of faces, before working on connecting them together with the rest of the cockpit. I needed to keep an eye on where mesh flow would be tighter to act as holding edges and where to even out the geometry for a smooth finish. Unfortunately. My final topology ended up with a lot more geometry than I was hoping for. This made it harder to keep smooth, but using edge flow reduction methods on a non-flat hard surface can lead to artifacting, so I had to make do.



Finally though, I managed to get the retopology completed. I could now get rid of my live object and add thickness to the cockpit. To do this, I simply selected every face and extruded inwards. Now, Maya’s extrude tool can result in overlapping polys due to how it scales vertices, so I had to redraw a few edge loops on the inner side to clean things up. Finally, I closed the shape with a back wall that would separate the cockpit from the rest of the helicopter’s cabin. Here, I’m not worrying at all about mesh flow since it’s just a flat surface. And here is the final cockpit! I have to say I’m quite happy with it. Finally.



Landing Skids

Now, I could finish the landing skids and add details. This part was quite easy, since it’s just a bunch of tubes. I simply stretched a bunch of cylinders to get the job done. I made use of the wedge tool in Maya as well, which lets you easily extrude faces in an arch shape. This made creating the curves in the landing skids significantly easier.


The devil is in the details with the landing skids. I didn’t need to worry much about edge flow in this part, since these pieces are tiny and will barely be seen by the camera. Still, I took the opportunity to practice creating some complex shapes.


Here are the final landing skids!



Details and Cleanup

Now, all that was left was detailing the model and cleaning up any messy areas. I added in those large blades in front of the rotors, a pair of guns by the cockpit, and some connection pieces.


I also added a few tiny detail pieces here and there. After all the detail work was complete. I went through the entire model and cleaned up any geo that looked unclean. I even found a component that I had skipped adding holding edge loops to! This particular piece was giving me a lot of trouble when it came to adding holding edges while maintaining clean geo, so I ended up splitting the piece into two.


I also took my old base shape for the cockpit before I smoothed it (recall I keep a backup folder for parts I may need to go back to. Convenient!), and scaled it inwards to use as windows. You can "scale along normals" in Maya by selecting all of a mesh's vertices, then using the move tool set to "normals" to move each vertex inwards along its normal.


Finally though, my 3D model was completed! Of course, I could probably go back in and add even more details, but I needed to know when to stop since the main priority for this shot is the FX (the Samson is quite far from the camera too, and on fire, so it’s not like extra details would be noticed)


Before stepping away though, I took the time to properly rename all the pieces in my scene, and to clean up the scene outliner. This is always a good practice to have.




Importing the Model and Resimming

It’s time to import the model! This was pretty easy, another file node to bring in the .fbx file and then making any needed adjustments (I had some normals flip on import). For the collider geometry, I selected the rotor blades and the bars between the windows and added a peak node to thicken them, so that they are more easily read to collide with.


Then, it was all about going through and resimulating FX elements to accommodate the new model. This wasn’t necessary for each one though. For the initial fire, I tried to create a new source, but I was encountering issues with blobby sources again. I had an idea of what to troubleshoot for this, but for the sake of time I’ll be choosing to just stick with the old simulation moving forward, since the general shape of the helicopter barely changes between the proxy and hires in that particular area. So, using the old simulation is hardly a problem.


The explosion was easy. Only thing changing is collisions (and even that barely made a difference)


For the main pyro, I did have to repaint parts of my source to match the new model, especially since the shape of the cockpit changes much more with the new model. (It’s a bit more squashed).


For the simulation itself, I also decreased the amount of wind and increased the power of the velocity field to try and get rid of that vertical motion from last week. I also greatly reduced the amount of downwards force in the field. One final change I made was to mask out the velocity in the rotor area where the field was, so the velocity noise placed on the original source would not interfere with the velocity field.


At this point, I also realized that I wasn’t merging the velocity between both sources properly. Simply using the merge node create two v inputs that confuses the pyro solver and may lead to strange results (thank you to my professor to pointing this out to me when we met up during the week). He suggested simply renaming the v attribute to v2 and then rasterizing it as that. Then, in the pyro solver, I could add a new source called v2 and convert that to vel.


Finally though, I could resim the main pyro.


The Final Render Flipbook

Unfortunately, I completely ran out of time to render this week, so I had to stick to a flipbook of all my elements this week.



A few notes: The fireballs have a major stepping issue. You can see the blobby source on the initial fire, but that’ll be replaced with the old simulation, since I realize it wont make much difference are the shape of the proxy is similar enough to the final model. The main pyro looks more promising, but will have to be toned down significantly, as right now it’s resulting in huge swirls that are moving way too fast. Also, pyro trails are not in this shot as I have begun trying to fix the problems in them from last week, but didn't get far enough yet to show the changes here.


But, that’ll all be for next week. See ya!

3 views0 comments
bottom of page