Modelling

Week 8: Intro to Modelling with Blender by Valzorra

After reading week was over, it was quite the relief to get back into my normal working schedule and to begin creating assets and prototypes during Week 8. At this stage, I had completed my idea generation and had come up with four fleshed out project proposals, one of which I was thinking of taking forward. Aside from mathematics, the common element between all four project proposals is that they would work best in a 3D environment. There is a potential problem with that as I have never really interacted with any 3D Modelling Software, aside from brief endeavours into texturing during Year 2. However, I am really passionate and enthusiastic about these projects, I want to bring them to life and to make them happen, which is why I have chosen to take on this challenge and learn how to create 3D Models. Week 8 was the perfect time to do this, because if I couldn’t handle the process at this stage I would have enough time to reconsider my options and potentially join a team. Nonetheless, I was determined to try my best and to create an asset by the end of the week. My program of choice was Blender as it is a professional-grade 3D Modelling Software, free of charge, and it was the program I used to texture last year, which meant I was conversational with its interface.

I started my journey into 3D Modelling with Blender by attempting to create a low-poly proportional female figure. Zeta, the protagonist of Project Proposal 4, will most likely be a healthy and athletic female, which is why I thought that trying to create a body for her as practice may be useful. Additionally, I have chosen to focus on a low-poly style as I am only beginning to learn 3D Modelling, and focusing on stylised and low-poly models would be a more feasible choice for my FMP. Overall, I would like to reiterate that my main goal here is to become familiar with Blender, to learn some of the basics of 3D Modelling, and to determine whether or not this was something I thought I could handle if I were to work by myself. The finished asset itself may not necessarily be used in the game, but if it does make it’s way to it, then all the better. Now, without further ado, let’s dive in.

Make Human

Before I started working in Blender, I thought I would explore other software for crafting Human 3D Models in order to create something to use as a reference for my own model. James S. recommended that I explore Make Human as it’s an open-source program designed specifically for making anatomically correct human models. Upon downloading and opening the program, I was surprised at how easy and user-friendly it was. Make Human generates a human 3D Model and allows users to change their gender, shape, form, and more, through the use of sliders. One can make humans taller, shorter, more muscular, chubbier, and all facial features may also be amended. Make Human even has a few in-built hair styles and clothing for the model, all textured. Below is an example of the default human model and the way the sliders work in this software.

The whole process of crafting a model was really easy and straightforward, and all I did to create my character was to work with the default model and to adjust the sliders accordingly based on the features I wanted to change. I made the woman a bit slimmer, more muscular, and taller than average, as she is meant to be quite healthy and relatively athletic. I made her face slightly sharper and her nose a bit longer than the default model. After I was happy with my character in terms of facial features and overall body shaped, I took advantage of Make Human’s built in poses and facial expressions to bring more life to the model. The image below shows the finished character with Make Human’s textures. I went for a braided hairstyle as it seemed to work best with her face, and a pose to make the model seem more life-like. However, I would like to reiterate that these decisions are not particularly significant just yet, as I am merely using this model as a reference by which to create my own 3D Human Figure. Once I was happy with the model and it looked somewhat like what I envisioned for Zeta, I exported it as an FBX (as Unity tends to like that format), and was ready to begin working with it in Blender.

Modelling in Blender

Once I had my FBX from Make Human I imported it into Blender to use as a guide for my own model. I chose to work with a model in the A-Pose, as my character will be rather unlikely to lift her hands up past that point, allowing me to model the shoulders with more detail. This is in contrast to the T-pose, where detail on the shoulders can get lost or distorted while doing animations. They key bit to remember here is that I am making a low-poly yet proportional human, which is why the highly detailed Make Human (MH) model serves as nothing more than a guide. I started off with a simple cube at the bottom of the left foot and I began shaping it and forming it along the outline of the MH model in order to get an abstracted version of the foot. One of the main tools I used was the Extrusion Tool, which works by selecting the face one would like to extrude and then pressing E. What’s lovely about Blender is that it automatically determines what type of extrusion would best fit the selected face, thus making the process a little easier and more user friendly. Another tool I used very commonly was the Loop Subdivide Tool (Ctrl+R), which creates a loop of edges and vertices around the selected object. This was hugely beneficial as it’s the way I was able to divide the cube I started with into a shape with more than eight vertices. Once you have those vertices, they can be dragged along each axis and changed into desired shapes, which is how I formed the foot.

The original imported A-Pose Model from Make Human.

As I was working with my cube, continuously using the Extrude Tool and the Loop Subdivide Tool, the model I had was slowly starting to look like a foot with an ankle. The MH Model was incredibly useful as it gave me insight into where to break the lines and form the appropriate curves of the human body. In that same fashion of using these tools, editing my vertices, shifting my edges, and occasionally grabbing entire faces of the model, I went up along the leg and was sure to include the key curves and breaks in it. I really enjoyed this process of abstraction because I had to continuously ask myself how much could I take away from the shape of the foot before it ceases to resemble a foot anymore. I was aiming for a relatively abstracted version, but still clearly distinguishable as a human foot and leg.

Going along the leg and progressing towards the waist with the described techniques.

Up until this point, I had been working only on one side of the leg, slowly creating half of the body. After a handy tip from Richard, I turned on the Mirror Modifier in Blender, in order to connect simultaneously create both the left and the ride side of the human. The model was mirrored along the X-Axis, with Clipping and Merge enabled, as I wanted both sides of the model to be connected and to result in one coherent shape. After enabling that modifier, I continued to go up along the body and to extrude it all the way up, working along its curvature. The most difficult and and time consuming areas thus far were by far the areas of her bottom and her breasts, as those we most shaped like an oval, which means they required significantly more subdivisions of the plane. Nonetheless, I was greatly enjoying the process thus far and was excited to complete the finished piece.

Once I had crafted the body up until the neck, I branched off to the sides to complete the arms. Using the same methods of Extrusion and Subdivision, I proceeded to form the elbows, the muscles, and the armpits, which were rather challenging. I had the most difficulty viewing the area underneath the armpits as positioning the camera in a convenient way to edit the vertices often caused me to see the inside of the model, which was not very helpful. Nonetheless, I managed to work my way around that issue by positioning my view towards the side of the model rather than directly beneath. Another very challenging part of this area for me was the hand. I knew that I wasn’t going to create all five fingers as that seemed unnecessary for a low-poly model, so I focused on the thumb and combined all four fingers to essentially create a mitten. Nonetheless, I found it really difficult to account for the curvature of the palm, and I required a series of subdivisions to get that done correctly. After it was all complete, I was rather pleased with the result as it did resemble an abstracted human hand. However, I am eternally grateful I do not need to repeat the process on the other side.

Extrusion of of the area around the armpit, which will eventually form the arm.

The hand and arm near completion, by far one of the trickiest areas to get right.

After I had created the entire body, including the arms, it was time to proceed to the head, the section I found the most challenging. In the same fashion as before, I began by extruding a small section from the neck and slowly matching all of my vertices to points in the face where they touched the MH Model. The difficulty with the head came from the fact that it was the most ovular part of the entire body and thus required more vertices to get it done accurately. Additionally, as this is a low-poly model I did not model the face based on the MH Model, but rather made my own version, which only had very basic indents to indicate the eyes and the mouth, and a simple bump for the nose. I did go over and edit the face a couple of times because I kept making mistakes with stray vertices and geometry, but nonetheless, I got there in the end.

Blender25.png

Below is an example of how I would fix any mistakes I found within the face and the entire body. Whenever I notices that something was off with the model in Solid View, I became suspicious of what has transpired. To investigate the issue, I would enter the Wire Frame View of the model, which only displays the mesh of the model. Once a face is selected as shown below, if there are any stay vertices or edges, they become really easy to spot. The way I fixed problems of the sort was a bit clumsy, but nonetheless, efficient. I would merely delete or dissolve the problematic vertex and then merely fill in any holes the process may have caused in the model. I was rather lucky, because I did not have an abundance of such instances at all, so there was no need for excessive deletion and rebuilding. Once the head was all done and I had fixed all of my mistakes along the way, I went on to clean up the model from any stray or unnecessary edges and vertices, which significantly decreased my Vert Count and my Tri Count, making the model much more optimal.

Blender28.png

Once I was happy with the model overall, I took a look at it from afar and tried to experiment with some of the Modifiers in Blender. Specifically, I wanted to make the model look even more low-poly than it already was. That is why I explored the Decimate Modifier, which essentially decreases the number of Polygons in the model. I found the process rather exciting because by editing the Collapse Ratio, which is a slider, I could change the model from its current form to a single triangle. The question I was asking myself while using this Modifier was how much information can I take away before the object is virtually indistinguishable. I found that for now, the ideal ratio was somewhere around 0.3050, however, I could decimate it even further if needs be. Another Modifier I explored was the Triangulate Modifier, which essentially divides each face of the model from Quads to Tris. I personally much prefer Tris, because mathematically, any three points can fall into the same plane, which is quite nice to model and work with. Quads on the other hand, do not always have all of their vertices in the same place, which can lead to some difficult to spot and undesirable geometry. Once all of the Modifiers were applied, the model was complete.

Blender39.png

Thoughts and Reflection

To my surprise, I’ve really enjoyed creating this model. The process was not overly complex and it was an excellent exercise in familiarising myself with the Blender, some of its shortcuts and its interface. There were plenty of challenges along the way, most notably with the more ovular shapes of the human body, but they were all manageable and I managed to work through them without too much trouble. One very frustrating setback was that while I was crafting the head, my computer completely crashed causing me to lose some of my progress. However, I took it as a learning moment, and acknowledges that I should be saving rather frequently as there is a risk of things like that happening. Through this model, I have now familiarised myself with Modelling via Extrusion, working with vertices, edges, faces, I’ve explored a few modifiers, and I have clearly pushed the technical limitations of my personal computer. I’ve explored how to fix certain problems and how to optimise my model by dissolving unnecessary geometry. To top it all off, I am rather happy with the final result, which is attached below. This is my very first 3D Model and it seems to look rather decent, better than I expected in any case. The model seems fairly well optimised with 934 Verts, 1,920 faces, and 1,920 Tris, although there does not seem to be a solid number one should aim for with these stats. Creating this figure took me about three days during which I was also working on other prototype pieces. This should hopefully mean that if I stick with a relatively low-poly style, I should be able to feasibly create a 3D game by the end of the year, which is incredibly exciting. This exploration has been incredibly motivating and I am looking forward to getting back into it.

HumanBody.png