Hello hello! I am now 4 (or 3, excluding the proposal) weeks into my shot recreation! Let's get right to it!
More Layout Work
Before we get into the fx side of things, I got some feedback on my layout that the islands still felt to close to the camera. So, this time I just went crazy with the scale of the islands, making them super big and pushing them really far back. My professor did say I shouldn't be afraid to use all that empty space and so I did! I had to tweak my camera and helicopter animations just a bit to fit things better, but in the end I got somewhere that I think looks better.
And of course, I imported my updated Samson proxy into the scene. Just a quick note that I might not get much modeling done before I post this blog on Friday, but I am ready to begin working on the high resolution model at this point.
Here is a flipbook with the adjusted layout:
Fixing the Initial Smoke & Fire Emitters
The next item on my list is to adjust the emitters for my initial smoke and fire elements. Right now it's just a selection of faces on the helicopter, and as such on the you can clearly see the blockiness of the emitter at the start points of the smoke and fire. So, it was time to create a new emitter. And, since I am importing a new proxy too, this gives me the chance to start fresh.
Emitter Setup
I'm creating the basic emitter setup in a separate node now so I can do an object merge for both smoke and fire, rather than having to do things twice. To start, I selected the main faces on the Samson that encompass the area that is on fire, put them on a separate group, and blast the rest. Then, I scattered a whole bunch of points over the area. My plan is to paint patchy areas on the block I have that are on fire, have those areas fade out a bit, and then tie my density to the painted areas. Sorry if my wording there is a bit weird, but it'll make sense! Luckily, Houdini has the perfect node for me: Attribute Paint. I created an attribute called "emit" set to 0 (I could have used density instead, but I wanted to keep thinks separate just in case), and then used the attribute paint node to literally paint on areas where I wanted the "emit" attribute to go up to 1. The node also gives the option to adjust the brush's "Soft Edge", which produces a falloff as the emit attribute fades from 1 to 0. Also, I set the brush shape to "Screen", which let me paint the particles inside the brush circle in screen space.
And that is my setup for the emitter! Next, I'm going to apply it to my smoke.
Adjusting Initial Smoke
I started by using an object merge to connect my emitter. Then, using an attribute wrangle, I created the density and temperature attributes and set their values to equal the emit value. I decided I wanted to adjust the falloff a bit more, so I dropped in a pair of attribute remap nodes--one for density and one for temperature--then adjusted the ramps to my liking.
After that, all I really had to do was connect the new emitter to the rest of my setup. I changed the pyro source mode to keep input. Then, I adjusted the noise I had on the density and temperature a bit, and messed around some more in the pyro solver settings, until I got a look I liked.
Here's the new smoke. It looks a bit puffier(?) now, I'm not quite sure how I feel about that. I'll get my professor's and peer's opinions on that:
I might still adjust the emitter in the future, but for now I'm happy with what I have.
Adjusting Initial Fire
Once the smoke was done, I had to repeat the process for my initial fire. This was slightly different, as if you recall I did do a different type of setup for my flames.
I started off by just cleaning up my original setup. I did have a bunch of switch nodes where I experimented with different setups last week, and I don't need those anymore. Then, I imported my emitter. This time, instead of using a wrangle to bring in density and temperature, I skipped to directly adjusting my emit attribute using the attribute remap node. After that was the trail node, then a another attribute paint so I could add some small difference to the emit values in the trailed pieces, setting the paint mode to multiply. The reason I didn't simply use an attribute noise is because I didn't want to affect the emit values on the main piece. Finally, I blasted out all points below a certain emit value. 0.4 worked for me.
Now moving on to pyro. Once again I put in a pyro source with mode set to keep input. Then, I used a VDB from particles to get my volume source, same as last time. This node gives us a density, and the temperature is added by renaming the density attribute to temperature then merging both together. Finally, I added volume noise to both attributes before connecting them to my existing DOP network.
Now, when I did this, I spotted a very obvious cut off of the flames where my emitter particles were blasted:
So, how to fix? Now I know the proper method is to have the density fall off away from the points, but unfortunately I could quite get this to work. I wanted a way to transfer my emit point attribute to the density volume attribute, but I couldn't find a way to do something like that, hence the blast node from earlier. So, I went back to my emitter, just before the blast node, and added an attribute noise node to add to my emit values. I set the group to @name<0.05 to only add to the parts where there is very little emit. Then, a used a second attribute noise set to multiply to get a few points in those low emit areas where the emit attribute goes higher.
Now, when I go to blast low emit points, I get this (note the extra scattered points):
So now there wont be as much of a clear cutoff of density in my volume. It's not an ideal solution, of course, but it's my temporary workaround while I try to find something more efficient.
Now, one last thing, my flames needed more chaos. So, I turned down the dissipation slightly and added a lot more noise to the flames, and a bit more to my emitter too.
Now, it's looking a bit patchy due to how I set up my emitter. I'm not a fan of it, so I'm going to try doing the volume rasterize and pyro solver method again instead of the DOP network method, just to test. If this renders correctly this time around too, then great. Since this removes the need for a VDB from particles, I get more control over my density and temperature falloffs directly from point attributes. Here is the result of this:
It does look better, so I'll see how it renders and decide from there. (Spoiler: it renders! Which means I'm going back to using the pyro solver instead of DOPs)
Things go Boom
Now we're getting to the part that's actually relevant to the title!
To start, I'm going to do a bit more research on pyro explosions in Houdini. I'm going to try and familiarize myself more with the pyro burst source node as well as the built-in aerial explosion configuration.
Time skip! A major thing I caught while studying the aerial explosion configuration is that I had done my pyro bake shader wrong, hence why my explosion was looking like it had no smoke in the render. I had the "Fire" section checked on in my pyro bake where I should have had "Scatter" instead. I studied the SideFX documentation of the pyro bake node and the scatter section and made adjustments to my explosion that fit my liking.
Now the smoke is better visible in render!
I also made some small changes to settings in the pyro burst source node. A found this video that does a deep dive of the node and was very helpful. I would like to further art direct the shape of my explosion in the future though.
One more thing before I go to render: I keyed the scatter intensity value to decrease over time, in order to match the speed at which the luminosity of the explosion dies out in the reference. Same with the density of the smoke.
A Temporary Fix for the Holdout Matte Issue
So, remember the holdout matte issue I was having for the past couple weeks where it wasn't working? I think I might have a temporary fix for it, let's see if it works.
So, the plan is, import the Samson and missile (the only two assets that need to be matted) again, and this time assign a constant black shader to them. Then in Nuke I can get rid of the black areas which act as the matte.
However, out of curiosity I plugged in these new matte assets into my render geometry settings as matte objects and...surprisingly it worked. So, apparently something is wrong with my original Samson and missile imports that aren't allowing them to be matted out, but hey at least something works now!
The Final Render
...still can't get my missile trail to render. I've made some changes but not enough to really write about here. I'll have to mess around with it more, and I'll cover whatever changes/fixes I made in next week's blog post. I hope to have it render-ready by then. I did also get started on the fire that continues after the explosion, but it looks really terrible right now and I don't want to render it at this point (will also cover that more in next week's post).
Also, I rendered the initial smoke and fire together right now because they look really weird when rendered separately at the moment.
Here is my final render for this week:
Comments