This weeks’ session the task was to continue destroying the building from the previous session, with the meteor animation Mehdi prepared for us. I had a few issues on this task and keep having them, each time I fix the problem with Mehdi on a 1 to 1 and I face another issue not much time after. Houdini is complex and these sessions are great to have a better idea of it, it needs solving every time.

We began by assigning active points and static points, as the active points would be the ones affected by the destruction and the static would be the ones who stay in its’ place. For this we created a simple geo of 5 spheres and used a motion trail node to see preciselly the path of the meteor, group it and with an assemble node create the packed geometry.

The for each nodes are used to run a number of loops inside a simulation and the time it takes to complete this operation depends also of how many cores your cpu has available, not all nodes are able to run in this method but the ones that are can be fastened by this process.

After running the compile block with the for each nodes we got some artifacts, to solve this we set up an expression to change the method on which the operation is being run, for this we create a switch node and write an expression on it.

When we are happy with the setup and there are no more artifacts we create a file cache node and save it as part01, this is also a way to make the simulation run faster as it is now running from disk and not calculating all the nodes again. We repeat this file cache process for the concrete nodes – facture of the building we did in the previous session, save it as part02.

For the glass setup, we create a for each block and copy the group made for the factured pieces, this way we know what glass we need to break, add an rbd material facture and run the compile block. We had some artifacts, and I had to set another metting with mehdi to solve this as I was having more issues than mehdi in the tutorial. We fixed it, the problem was that some glasses were duplicated and the facture was creating artifacts. After this was solved and we were happy with the result, we repeat the file cache node process and save it as part03. The same is done as part 04 and part 05 for the bricks and aircon.

Finally we create a null, a vdbfrompolygons and a scatter and a voronoifracture node and the static points cache is also created, saved as part01_static. This is how the complete setup nodes look like:

When this process is complete we merge all the geo we just setup, create packed geo and an rdbbulletsolver node, that will be used later, first we need to create another for each and compile block nodes with a polyreduce node and a switch with an expression to make the simulation faster. Both these geo need an assemble node to be made in to packed geo and we can run the simulation. The result is that the whole building just destroys itself as all we’ve been doing is destroying it.

Next, we will copy the active group set up and set another expression with the attribute wrangle node and we run the simulation. This is how it looks like:

To create an explosion and some drebi projected on to the air we can use the pyro trail path nodes which will create a small animation for the debri to fall and all this can be adjusted on the parameter window.

This will need to be placed on the right place and the right frame, in this case we set this trail path to 6 frames and got another one on the end of the building, where the meteor comes out. This is the result of the first debri explosion without the fire simulation:

This is how the geo RBD SETUP nodes looked like in the end:

This week I sketched the storyboard and built up the 2 master file scenes from the cafe and the store. I looked up some free environments online but none of them was good so i built them up with assets I found on turbosquid and cgtrader websites. The storyboard was drawn by hand as I left my pen tablet in london, the drawing is really just a sketch to have an idea of what shots will i need, with simple stickman drawings.

Storyboard:

After the storyboard I began creating the scenes. The store was the one that took more time as I had to set it all up, I modeled the wall with the showcase and door as well as door handle, in the end I added a few lights as a test, I might make a few changes later.

Interior of store:

For the exterior I made a basic set up and added a street lamp to be more aware of shadows and build up environment, using the same hdri image for both environments.

For the cafe, I used the corner street environment available on the resource share and added some assets to create some sort of “terrace” from the cafe on the street. I deleted the windows from the ground floor of the main building so i could add a material to look like window glass and some reflection. The only light source of this scene is the skydome light with the hdri. This is how the cafe scene looks:

When testing the materials and light with an arnold render i noticed that my character, amanda has some sort of light attached to her so I’m still trying to figure out how to solve that.

For this next term we have to make at least 2 projects one solo and one collaborative project. This is the part where the masters really starts to change its shape into what we want professionally so the next step of our education is self-directed study, to be able to work on our own and plan it properly. Ofcourse, in a stage where I feel unprepared being told what to do is more comfortable, but that is the point: to take us out of our comfort zone and learn from it.

My proposal for this project was an animated short story, with a story of my own, although that seemed over ambitious it’s never too bad to try it. So I began writting a story, and another and realised that I had a huge difficulty in finishing the story, what do I want to tell? What is the goal of all that? But I eventually wrote one, too big in fact. I set up a 1 to 1 with Luke and realised that i could do the story i wrote for my FMP and for this term’s solo project I chose to do a short story but this time, really short, not trying to explain anything just using a performance to tell a story. This way I believe I can improve the body language of my animations.

This story is about greed and fake friendships. A girl sees a handbag and immeadiately falls in love with it on her way to meet her girl friend for a tea, she walks in the shop, asks for the price and leaves as she has no money to afford it. When her friend shows up, she has the exact same bag she just saw and asks if she likes it, jealous of her friend, they drink their tea and a ninja comes and kidnaps her friend, surprised the girl sees her friend being taken, looks at the bag and keeps it to herself.

When searching for rigs I found a really cool website that had a lot of them and some free! So, the 3 characters I will use are from that website, Amanda, merry and Ninja.

The environments I need would be a cafe or a bar, the hall of a store and a street where the store is. I’m still looking for good environments but I can also create a simple environment with some assets and set a scene.

The second performance animation we had to make had to be with two characters in which at least one had to speak. So the first task was to search for an audio clip and have it approved by luke, then look for rigs from the asset library, record reference footage twice and start working.

The first 3 audio clips I chose were from Final Space (netflix series), The hustle (movie) and from Peaky blinders. The first one couldn’t be as it is from an animation, the second was not very good (I admit) so it was the peaky blinders clip. It is a dialogue where Polly was speaking to Grace in the moment she founds out the betrayed Tommy. It is a slow dialogue but I believe it will be a good practice to know how to make characters move without any major movements. I do have some trouble with timings and this seemed like a good exercise for that. Although I went looking for more audio as I regreted choosing this one, but I ended up keeping it.

The characters I chose were Janine and Lou

I had Janine as Polly and Lou as Grace and began recording the footage. The first one was too bad so, I re-recorded it and the result was much better.

I chose the environment “Victorian Interior”, found a table, glasses and a fancy botle on turbosquid and built up the scene. I began by animating phonemes from each character and then animate the body according to my reference footage, Polly was first then Grace.

These two characters have to pick up and change hands when grabbing objects and that, I believe, was where I lost and keep losing time with as I was using the animSnap tool to have the objects constrain to the hands but they would go crazy and the objects just appeared on the other side of the scene. My solution for this was to block everything like they had the objects on their hands and in the end I would deal with it. I ended up only parenting one cup to grace’s hand, as she wouldn’t have to pick up anything else with that hand. This is a problem I will have to learn how to fix and really understand constraints and all that comes with it.

This is the final playlast of the blocking:

I still have to figure out how to light this scene properly, this environment came with a ton of lights but I would like to improve on it and be the one to light my work as it is such an important concept of emotion and reality.

After the blocking I tried to make a shot sequence with the camera sequencer and this was the result:

This second session of lighting was made with katana, a software by Foundry and in my opinion was better to understand the concepts KK was teaching us. We used frames from the movie Fight Club (1999), by David Fincher and lighted a CG scene with the same lights as the frames from the movie. Nuke and Katana are very similiar but I think Katana is bit more intuitive and easy to understand.

KK prepared a file for us and gave an intro at the beggining of the session which gave us a proper workflow on how to work with multiple shots like organizing the shots by angle and re-using light rigs. The angle of the shots and the amount of light they get can sometimes change but that can be tweaked in katana when re-using the same light rig with a small variation for a different shot.

One of the things KK mentioned was when looking for an HDRI an important aspect to be aware of is the position of the sun, as the HDRI can be used as a light and on the website HDRIHaven they preview what the light looks like on different materials which is extremelly helpfull.

The first step of this exercise was to set the graph state variables (GSV), change it’s name to “shot” and add a variable for each shot, in this case we created 5 variables named “shot01, shot02… etc”. After this we go to the node editor and on the lighting node group we can see there is a Gaffer Three node this is the master light rig which then will be tweaked and changed for each shot or camera angle and overwrite them accordingly.

For this we create a Variable Switch node and add it bellow the gaffer three node, to add more inputs to this node we just need to hit the arrow until we have the desired number, in this case we created 4 inputs and connected the gaffer three node on those 4 inputs we just made. To see the parameters of a node we need to either hit E when that node is sellected or hit the square on the right side of it until a green color appears. On the right side pannel we name the variable as “shot” and in the patterns we can set 1, 2 and 3 for each shot group. So the first will be shot01 shot03, the second will be shot02 shot04 and the last one will be shot05.

Before changing anything else we made a few renders to be sure of how they work on the setup KK made for us. A render for just the characters and everything together (background and characters) and these can be seen on the catalog tab where we can see a list of the renders we made.

Back on the node graph, on the input one we create a new gaffer three node and connect it to the according connector. On the parameters tab we have no lights but if we hit “show incoming scene” we will see what the above gaffer three node has inside on this node and change it just for this input. We deactivated the fill light so we can focus only on the env light, to edit it, right click and hit adopt for edditing. With this we can see new parameters showing up on the pannel bellow in the material tab.

To activate live renders for the lights we are using we go to the scene graph, that works like maya’s outliner and we can see everything this project has, we select the light group and check the live render box on the env light so we can have it rendered along with the image. On the object tab we can apply rotation to this light and match it to the right setting according to the movie frame we are trying to replicate while in live rendering. Hitting “S” to change between cg render and original frame.

This process was only for shot 1 so now we have to check if shot 3 (that is on the same camera angle group) is correct with this setting. As it isn’t, we have to add another input on the variable switch node and connect the shot 1 gaffer three node to this last input and get shot 3 on this last input. Add another gaffer three node on input 4 and with a live render tweak it’s rotation to match the original frame. This is how it looked like after tweaking:

After the light matches the original frames we can make a render of the prepared plates KK made for us, where there is only the background of the original frame without the actors and check if the light matches. The exercise for this session is to match every character with the original plates, so we must do this for all the frames.

When comparing the character render with the prepared plate we can select “toogle overlay/underlay controls”, middle mouse drag the background image to the underlay spot and we have the final image for this shot.

Shot 1:

Shot 3:

KK told us to do the rest of the shots with what he had just taught us.

Shot 2:

Shot 4:

Shot 5:

The only thing that I was unable to find out how to do was to export the render images, everytime I rendered to disk nothing appeared so I took screenshots of the final frames. I am aware that some things are wrong and I expect to review them with KK on our 1 to 1 next week.

This week’s session was about experimenting some examples of particle effects mixed with desintegration, like we did with the particles but this time using the voronoi fracture where the destructed pieces would flow and disappear like the particle simulation giving it a more magical effect.

Desintegration test

So the first try was with a torus shape and with the scatter and voronoi fracture process that we have been doing but this time with a DOP network. Inside the DOP network we used a pop force and a RBD packed object and a bullet RBD solver connected to a multisolver node – usefull for when working with multiple different solvers like this example. But this only allowed us to desintegrate the torus all together and what we want to do is this effect happening from right to left so we had to create a group, assign it to points and check the keep in bounding regions option, this way a bounding box with be created and it can be keyed and give the result we want.

This group we just created is called static and now we need an attribute create node to tell the software that when the bounding box is out of that region the pieces of the fracture can have the magical effect created in the DOP net.

I had an issue when performing this part of assigning active and static parameters because I put the name active in caps lock like the static and only when I redid this part of the exercise I realised that the name active really had to be in lowercase. In the end we applied this effect to a test geometry tommy.

This is what the nodes looked like:

Flipbook:

Smoke with particles simulation

The next task was to make a smoke simulation with particles and have them react to the smoke movement. So the first step was to create a circle, scatter points from it, a pyro source node and a volume rasterize attribute node to assign density and temperature. Also attribute noise from density and a pyro solver to have smoke, like we did in the previous session.

After adding lights and tweaking the smoke simulation we can convert to vdb and cache it to disk with a file cache node. When this is done we can take care of the particles with a DOP network and scatter the points from a circle again. This is what smoke simulation nodes looked like:

This is what the DOP network nodes looked like:

This was the final result:

Big Destruction

The final task of this session was more complex as it had imported geometry and animation, Mehdi prepared a model of a building and a meteor simulation so that we could do a big destruction on the building with the meteor. So the first step was to import the assets and clean it and see if the geometry was prepared for destruction as every shape needs to be closed and can’t have artifacts (errors in geo).

Before all that we need to check where the destruction is going to happen, after that we need to clean the geometry at least from those parts, the first was the floor from the different floors of the building, then the entrances, then the walls and then the columns.

Nodes after cleaning geo:

Now we need to clean the geo of the meteor, it’s geometry was too heavy for a rigid body simulation so we need to reduce it, for that we unpack it, convert it to vdb and with a time shift node we can remove the time dependency and have the proxy calculated only in the first frame, this way the animation is much lighter and simulation runs faster. After cleaning the geo this is what the nodes looked like:

Now we need to destroy the building, for this we tried first with 5 rectangular shapes and the goal was to destroy them all but with different fractures, but i had some issues with it when using the for each piece node as it is much more complex and heavy for my machine. In the end the attributes disapeared and I need to search for a solution, but this is what they looked like: