Notifications
Clear all
Topic starter
23/10/2023 7:35 pm
Hi everyone, my name is Josh, I’m a 3D artist originally from England but currently living in Amsterdam. My idea for this project is a little weird but I do imagine it will be an interesting model to create, I plan to make a very small “Bug”gy powered by snails… I knocked up a concept that I quite like, based on a tubular buggy chassis but with twig-based shapes that could be found in nature.
Step one is to create the twigs. I will be using Houdini for the vast majority of this project and generating as many things as I can procedurally. I am still a bit of a beginner in this program so this contest is a good push to lean some new things. I created a tool to create twigs from curves, it generates a low poly version and a high poly version complete with vertex colours. With an integrated python script the 2 versions are exported from Houdini, then batched through the Substance Automation Toolkit to bake the high res colours and height details onto the low poly mesh, resulting in customizable, completely procedural twigs for the chassis. There are a lot of assets that need to be created for this project so I will try to automate as much of the process as I can to make the computer do as much work as possible for me in the limited time available.
Here is a little insight into how it works, the most logical way to do this in my opinion was to always work in the same axis, Y in this case, build a straight stick based on the length of the input curve, then deform it back to the desired shape. This is done mostly using VDB’s and volumes, then the texturing is done with noises in VEX and some tri-planar projections in VOP networks. The whole thing runs in a loop so the 2-curve type sticks are made together as one mesh
One of the more challenging parts of this was to make sure the 2 sticks lined up with each other, there is a kind of bump that needs to be modelled on the first stick for where the second one grows from, so the orientation had to be pulled from the second line within the first loop, deforming the line and calculating the orient from the first point on the second curve did the trick here, I have avoided using quaternions in the past as I find them very confusing but this helped my to understand them a bit better
I created all the curves needed and batched everything overnight, there are 36 different sticks in total so this was a massive time saver. I added random seeds to a lot of the parameters in the tool so the colour and shape of each stick is unique.
Next up was the leaf springs, Similar approach, procedurally generate a high and low poly leaf, then bake the textures. I added some controls to bend the sides of the leaf but it is again generated on one axis and then path deformed to the input curve. These were again batched overnight
Now I needed to tie the sticks together, I had an idea that the car will be driven by a spider so spider silk will make a nice way to lash everything together. I originally considered using the vellum solver to create some shrinking ropes around the joints but this seems a little overkill in this situation, and procedural shapes will solve much faster. All we really need to do to create ties is to circumscribe a shape around the geometry. Houdini has a great shrink wrap node to do things like this so that was a good start. Again I made a tool to create the silk from an input circle in this case, the position and angle of the circle starts a loop which randomly rotates and moves on each iteration and wraps a bunch of lines around the geo
Some additional noises and wires inside the geo creates a nice web-like look
Serhii reacted
Topic starter
24/10/2023 9:03 am
Everything was Pieced together and textured, I'm happy with how it looks so far
Topic starter
12/11/2023 1:10 pm
Next up I need some snails to power the thing, I started by poly modelling some base meshes in 3DS Max for this, then moved them into zbrush to sculpt the details, the topology on the shell was tricky to keep as quads but I managed in the end.
Then in Zbrush I adjusted some of the proportions and sculpted details onto the shell and the scales on the snail body
I then exported the high and low res meshes into substance painter to bake the details into a height map and use that to additionally shade in some of the details. I made a few variants of the shell as I need 4, for variations in the snail body I'll probably just use a colour correction with some random expressions later in Houdini.
I plugged these textures into a v-ray material and made sure the SSS was behaving correctly, pretty happy with how they look for now, I can always refine it further later on
Serhii reacted
Topic starter
14/11/2023 2:31 pm
I also need a driver, I decided on a jumping spider in the end, lots of legs will help hold the frame and the little arms can be used later to add some reigns to the snails. I've also wanted to make a spider for a while since I saw the incredible Samsung advert so this is the perfect excuse to get started. This was modelled solely in zbrush, I started by blocking out the shape and basic anatomy with zspheres, dynameshing it and slowly adding the details, then when I was happy with the overall look I retopped with zremesher and began sculpting the finer details to bake in later on
I then took the low and high poly meshes into substance, baked the textures and created a fairly basic base colour, most of this will be covered in fur so no need to go crazy here
I made a quick dirty rig for this guy in Houdini so I could pose him later, Houdini is great for quick rigs as the bone capture harmonic node more or less eliminates the need for weight painting and I attached some IK controllers to the legs and arms. I then started on a groom, the default workflow for this is pretty choppy so I made a lot of custom nodes for this that kept things procedural and made it easy to add variation with a load of noise and random attributes, was a bit of a learning curve on this but I enjoyed making this, I don't get to work on characters at all in my day job
A little render of the fuzzy boi
Serhii reacted
Topic starter
19/11/2023 1:41 pm
Another update, I assembled the car and gave it a bit of a rig so I don't have to manually pose the snails or axles, it works by projecting points down onto the terrain and deforming a skeleton based on the returned positions so they follow the contours of the ground
this posed a little bit of a problem as it meant I couldn't use any v-ray displacement, the geometry needed to be available to me within Houdini. just subdividing the life out of it would have caused some memory usage problems so I came up with a way to subdivide the mesh based on the camera view. The subdivision sizes are controlled by the camera view and focal plane, so only what will be seen close up is heavily subdivided
Serhii reacted
22/11/2023 12:13 am
Beautiful progress, i can already tell that the final render would be definitely high end feature animated film quality !
just-joshing reacted
Topic starter
26/11/2023 4:35 pm
Thank you!!
some more progress, the deadline is close now so I have to get the environment sorted. I made some procedural assets for scattering, adding variations to some attributes by seeding them with the current frame number. Here is some moss made with l-systems
Some spores made from splines
some flowery mossy things again made with l-systems
and some mushrooms, there were a bit more complex but the general workflow was a lot like the sticks, create basic shapes, add fine details with vex/vops, convert to vdb's and combine everything, then remesh to a low and high res version and bake the vertex colours and height in the substance automation toolkit
For the vegetation I would have liked to have done this procedurally as well but the timeline is getting too short, I used some assets from quixel but there was some work needed to get them to render correctly in v-ray, in order for SSS to work correctly I need solid geometry, since the megascan collections are meant to be used in unreal it makes perfect sense that they are single sided. To fix this I threw all the leaves through a loop in Houdini to bake the opacity and displacement maps back into geometry and add thickness to all the leaves and stems where needed
I then started blocking out the scene and added a few rocks and a tree stump from the megascan collection, again it would be nice to have made these within Houdini but there just isn't time, also most of these will be covered in moss later so they do not need to be a work of art. For the assets close to the camera I used the same techniques to get the displacement maps into the geometry as the points I will generate for the scattering need the final geometry in order to get accurate positions.
when it came to scattering points for moss I used the surface normal to restrict point generation so it looks as if the moss is only growing on the horizontal surfaces available, also I added an attribute that would switch the colour to make the moss look dead, this would work by randomly selecting points but also favor points that were not close to neighboring points, so the dead moss was more isolated in parts of the environment not densely packed with moss
The rest of the foliage didn't really need this so that was simply scattered points instancing the geometry
I still have to create a little path from the mushrooms to act as a track of some sort, then texture a few remaining bits, tweak the lighting a little light, render it and comp everything
Serhii reacted
Topic starter
30/11/2023 7:52 pm
I have some renders now so time to wrap this project up, you won't see until I post the final shot but I have gone for a very fantasy style forest with lots of glowing spores and mushrooms, this wasn't made with one render but rather 8, I set up some takes in Houdini for this as I want to make a lot of volumetric passes and lighting that just isn't possible in one take. I also set up some quick dust trails with the pyro solver so I can add a bit of movement to the image. The texture on the ground near the camera is a little ropey but that will be out of focus in the final shot.
Everything was rendered as deep data, this makes things easier in Nuke as it means I don't have to render holdouts to merge the images. Also when dealing with volumetrics, deep compositing really is the way to work, especially since you can do things like generate and even defocus the fog, super cool. Here is a glimpse of the Nuke script
I'll post the final shot after it has been submitted
Serhii reacted
