A Guide to Building a 3D Game Character PDF

Elisabet Heikkilä A Guide to Building a 3D Game Character Bachelor of Business Administration Business Information Technology...

Elisabet Heikkilä A Guide to Building a 3D Game Character Bachelor of Business Administration Business Information Technology Spring 2017 TIIVISTELMÄ Tekijä(t): Heikkilä Elisabet Työn nimi: Ohjeistus 3D-pelihahmon tekoon Blenderissä Tutkintonimike: Tradenomi, Tietojenkäsittely Asiasanat: 3D-mallinnus, peligrafiikka, pelihahmo Opinnäytetyö käsittelee 3D-hahmon luomisen kaikkia vaiheita mallikuvien teosta aina valmiin 3D- mallin riggaukseen. Työn eteneminen selostetaan niin tarkasti, että 3D-mallinnuksesta kiinnostunut lukija voisi luoda oman hahmonsa kuvauksen perusteella. Opinnäytetyön päämääränä ei siis ole luoda viimeisteltyä pelihahmoa, vaan ohjeistaa 3D-hahmon luomisen jokainen vaihe esimerkin avulla. Opinnäytetyö on jaettu neljään osioon. Ensimmäinen osio kertoo 3D-hahmon luomisprosessin teo- riasta. Toinen osio selostaa teoriaa hyvän 3D-hahmon rakenteesta ja hyvistä käytänteistä mallin- nusprosessin aikana. Kolmas osio sisältää perusteet Blender-ohjelmiston käytöstä, jotta ohjelmaan tutustumatonkin lukija voisi halutessaan opetella käyttämään Blenderiä. Neljäs osio puolestaan sisältää itse käytännön osuuden, eli 3D-hahmon luomisen ohjeistuksen. Käytännön osio alkaa hahmon mallikuvien, eli 3D model sheetin, luomisesta ja antaa vinkkejä mal- likuvien luomiseen. Kappale keskittyy eri tapoihin pitää mallikuvat yhtenäisinä. Mallikuvien teon ohjeistuksen jälkeen aiheeksi nousee itse 3D-mallin kokoaminen. Työssä kuvaillaan, missä järjes- tyksessä hahmon ruumiinosat luodaan, mitä työkaluja käytetään ja miten osat lopulta kiinnitetään toisiinsa. Jokainen vaihe on havainnollistettu kuvin ja kuvatekstein, jotta ohjeistusta olisi helpompi seurata. Lisäksi käytännön osuus kertoo pikaisesti, miten valmiiseen 3D-hahmoon lisätään teks- tuurit. Lopuksi käytännön osuudessa käydään läpi 3D-hahmon rigin teko, jotta hahmo voitaisiin animoida myöhemmin. ABSTRACT Author(s): Heikkilä Elisabet Title of the Publication: A Guide to Building a 3D Game Character Degree Title: Bachelor of Business Administration, Business Information Technology Keywords: 3D modeling, video game art, game character, game development The thesis describes the entire process of creating a 3D game character from the creation of the reference material, the 3D model sheet, all the way to the creation of a rig for the final 3D model. The process is described in such detail that a reader interested in 3D modeling could create their own character alongside the thesis. The goal of the thesis is not the creation of a polished character model but rather to walk the reader through every step of the way the help of an example character. The thesis has been split into four main parts. The first part describes the process of creating a 3D game character in theory. The second part focuses on the structure of a good character model along with some tips on good practices to make the work more streamlined. The third part is a brief guide to the basics of the Blender program so even a reader who has never worked with the pro- gram could make use of this guide. The fourth part includes the final guide to creating a 3D char- acter. The guide begins with the creation of the 3D model sheet and gives tips to keeping the reference images consistent with each other. After the reference images are done, the next phase is the creation of the 3D model itself. The chapter describes in what order the character’s body parts are created, what tools are used and how the finished parts are combined. Each step is visualized with pictures from the example character. The guide also quickly shows how textures are added to the finished model. Lastly, the chapter walks through the creation of a rig for the 3D model so the character could later be animated. TABLE OF CONTENTS 1 INTRODUCTION............................................................................................... 1 2 THE PROCESS OF 3D CHARACTER CREATION.......................................... 2 2.1 3D Model Sheet.................................................................................... 2 2.2 3D Modeling......................................................................................... 4 2.3 UV Unwrapping and Texturing............................................................. 5 2.4 Rigging a Character for Animation....................................................... 6 3 THE STRUCTURE OF A GOOD CHARACTER MODEL.................................. 8 3.1 Polygon Count...................................................................................... 8 3.2 The Default Position........................................................................... 11 3.3 Clean topology................................................................................... 11 3.4 The structure of the joints................................................................... 13 3.5 Good practices................................................................................... 14 3.5.1 World Scale........................................................................... 15 3.5.2 Naming Conventions............................................................. 15 3.5.3 Version control...................................................................... 16 4 THE BASICS OF BLENDER........................................................................... 17 4.1 Setup.................................................................................................. 17 4.2 The interface...................................................................................... 19 4.3 Hotkeys.............................................................................................. 24 4.4 Modifiers............................................................................................. 25 5 3D CHARACTER CREATION STEP BY STEP............................................... 27 5.1 Creating the 3D model sheet.............................................................. 27 5.1.1 The front and back views...................................................... 28 5.1.2 The sides and other views..................................................... 32 5.1.3 Bringing the references to Blender........................................ 34 5.2 Creating the 3D Model....................................................................... 36 5.2.1 The torso............................................................................... 38 5.2.2 The limbs and joints.............................................................. 49 5.2.3 The feet and hands............................................................... 56  5.2.4 The hands............................................................................. 59 5.2.5 The head and face................................................................ 68 5.2.6 Clothes.................................................................................. 88 5.2.7 Hair....................................................................................... 96 5.3 Unwrapping...................................................................................... 104 5.4 Texturing.......................................................................................... 109 5.5 Rigging............................................................................................. 115 6 CONCLUSIONS............................................................................................ 128 REFERENCES................................................................................................. 130 ATTACHMENTS LIST OF SYMBOLS Vertex: A point where two or more lines or edges meet. For example the corners of a polygon. Edge: The line between two vertices. Polygon: In 3D modeling, the word polygon refers to the individual parts that form the surface of a 3D object (mesh). In games, 3D models should be formed of three or four sided polygons. Edge loop: A line of edges that circle around the model as an intact loop. Optimization: In 3D modeling, optimization refers to creating a model as light as possible for the game engine to run. Optimization includes for example keeping the polygon count of a model as low as possible. Layer: In 2D programs layers are like transparent sheets which can be drawn on without it affecting the contents of the other layers. Layers can be hidden, unhid- den and their order can be changed. 3D programs can also make use of layers. Rigging: The art of turning into a static 3D mesh into something that can be ani- mated. Rig / Armature / Skeleton: A system for animating 3D meshes. A rig is made by creating bones, linking them together and attaching the final product to an 3D ob- ject. The bones can then be moved to change the shape of the 3D mesh. BLENDER-SPECIFIC TERMINOLOGY: Object mode: A mode where new objects can be created or moved around. Edit mode: A mode where the created objects can be edited in more detail. Only object at a time can be opened in edit mode. If a new cube is created in Edit mode, it will be considered a part of the same object even if it would be separate in the 3D space. Scene: The 3D space where a model is created. 3D Cursor: A tool which determines where new objects and shapes are created. 3D cursor can be moved wherever it would be most useful. The cursor can be brought back to the middle of the scene with the “Shift + S  “Cursor to Center” command. Origin: An object’s root point. The origin functions as for example as an object’s middle point when using mirroring tools and as the object’s axis when rotating. By using the “Shift + Ctrl + Alt + C” combination while in Object mode, the active object’s origin can be moved either to the object’s center of mass or to the 3D Cursor’s current location. Pivot Point: The scene’s pivot center which determines how an object behaves when scaled or rotated. Changing the Pivot Point can for example change whether the object rotates around its own Origin or around the 3D Cursor.  1 1 INTRODUCTION The goal of this thesis is to create a step-by-step guide for creating a 3D character for a video game. The guide describes each step from the creation of the reference images to 3D modeling the character in Blender. The guide also explains how the finished 3D model can be unwrapped, textured and prepared for animation. Each step is explained in detail so the reader could create their own 3D character along- side the guide. Each action and quick key used in Blender is explained so the reader can replicate the process without external help. The guide is aimed at peo- ple who have at least some knowledge of how Blender works but a chapter on the very basics of Blender is included for those who have never tried the program before and only have background in 3D programs like 3DS Max or Maya. A character model will be created as an example but the focus of the thesis is not on creating a polished character. Instead, the goal is to explain the process so someone else could make their own character using the thesis as a guide. The thesis will explain how this one character model was created while pointing out areas where problems arose and how the said problems could possibly be avoided in future characters. The topic of the thesis is quite extensive, but the guide will be based on my topical seminar paper from late 2016. The seminar paper focused on the steps of creating a simple 3D character model based on a pre-existing 3D model sheet while this thesis builds upon and extends far beyond that. The information and images in the thesis will also be used to create compact study material for a 3D character crea- tion course for the Kajaani University of Applied Sciences.  2 2 THE PROCESS OF 3D CHARACTER CREATION Creating a 3D character is a greater undertaking than people often realize. In large studios, the job is usually divided among several artists due to the time it would take and the wide range of skills that the creation of a single character requires. In smaller projects, the character designs, 3D models and animation might be cre- ated by one person but this greatly limits the amount of characters that can be added to the game. Usually it’s better to have a different artist for each part of the pipeline. This way the job gets done faster as the workload is shared. For example, the character designer can work on a new character design while the modeler still works on the 3D version of the first character. In some cases, a team might try an approach where several artists create model sheets and create their own characters for the game from start to finish, but gen- erally it is better to have artists focus on a specific area of the pipeline. This way the team can ensure the art style of the production remains consistent. (Masters 2015) While smaller teams may not have the luxury of having enough artists to share the workload this way, allowing an artist to focus on one thing instead of having them work as a jack-of-all-trades makes the process more streamlined. As an example, we could use the character models of the game Overwatch by Blizzard Entertainment. According to the Overwatch team’s lead character artist’s ArtStation submissions, the work is usually shared in the following manner: one person created character concepts, another one the models, third artist handles the rigging and animation work is left for a fourth artist. Weapons were also often created by a separate artist. (Galand 2016) 2.1 3D Model Sheet While gorgeous concept paintings of characters work well for marketing purposes, they are less helpful from a 3D modeler’s perspective. When the character designs and 3D models are made by two different people, a single painting of a character  3 in a heroic pose leaves many questions about the design unanswered. If the mod- eler has not familiarized themselves with the reasons behind the character’s de- sign, they may accidentally misinterpret certain areas and thus the feel the char- acter gives might change. (Pluralsight 1, 2015) 3D model sheet is a character designer’s tool for conveying their vision to the modeler efficiently. A 3D model sheet displays the character in a neutral pose from several angles. (Image 1) Usually the character is shown at least from the front, back and side. There may also be other information on the sheet. For example, drawings that highlight certain details of the character or even photographs of dif- ferent materials that the character’s design would need. (Zagrobelna, 2014) Image 1. An example of a 3D model sheet with rough concept art alongside a front view, side view and a partial back view. A few extra details are shown with added images.  4 One thing that makes 3D model sheet different from usual concept art is the fact the subject is drawn in an orthographic view (Pluralsight 1, 2015). This means that a character’s each body part is pictured as it would appear without the distortion caused by perspective. In a picture with orthographic projection every part appears their actual size regardless of their distance from the viewer. Image 2 demon- strates the difference between perspective and orthographic projections. Image 2. The difference between perspective and orthographic projection. (Script Tutorials) The references can be imported to a 3D modeling software and used as a blueprint when modeling the character. This makes the job much easier for the modeler, especially if they are not too familiar with the anatomy of the creature being mod- eled. It also allows the modeler to focus on creating a model with good topology without needing to worry about getting the character’s body proportions right. In a way, a 3D model sheet is a 3D artist’s sketch for creating the final model. (Plu- ralsight 1, 2015) 2.2 3D Modeling After the 3D model sheet is done, the images can be imported to a 3D modeling program. The images can be displayed as background images over which the modeler can begin shaping the character. The model itself usually starts from a  5 simple shape like a cube. By adding loop cuts and moving the vertices around, the 3D modeler can create the character’s body. Limbs can be modified from separate cylinders that are finally attached to the body once they are ready. The head can either start from a cube or be crafted polygon-by-polygon per the provided refer- ence pictures. If the character or object is symmetrical, the modeler can use mirroring tools to create a perfectly symmetrical model. When a mirror modifier is active, every change done to one side of the model will be automatically done on both sides of the model’s middle line. This greatly speeds up the modeling process. Once the symmetrical version is done, the mirror modifier can be collapsed onto the model so both sides of the model can be edited individually if there are parts that need to be asymmetrical on the model. Any symmetrical parts are though best kept under the influence of the mirror modifier to ensure all future steps like texturing and rigging can be finished quickly and painlessly. 2.3 UV Unwrapping and Texturing Game characters rarely are made of a single material like clay or glass. In order to give a character features like skin, clothing and hair colour, texture maps are essential. But before these maps can be created and applied, the modeler needs to unwrap the UVs of the model. UV, or UVW, represent the different axes where the 3D model exists. UVW represent the XYZ axes respectively. (Autodesk, 2016) The first step of UV unwrapping a finished model is placing the seams. Like in real clothes and map projections, the seams are a way to break down a 3D object into 2D shapes. The modeler will choose edges on the model, mark them as seams and have the program place the cut-out shapes onto a UV layout. The UV layout can be exported from the program into a painting program like Photoshop and a 2D artist can paint the textures there. The finished textures will be brought back to the 3D program and applied to the character model. Still, 3D game characters and objects rarely only use one texture map that gives the object colour, also known as a diffuse map. The model’s surface can be made  6 more realistic by using different texture maps that can for example alter the model’s transparency (transparency map), shininess (specular map) or shape (normal, bump and displacement map). (Slick 2, 2016) When it comes to diffuse maps, they may be either made from photographs or painted from scratch. It depends on the game’s art style which of these methods should be used. A game with stylized characters, like World of Warcraft, work well with hand-painted textures while a more realistic game, like Call of Duty, would look more comical if the otherwise realistic people and environments would have painted textures. In image 3, the normally-realistic graphics of Witcher 3 have been changed to simple painted textures. Image 3: Giving simple textures to a game with realistic models can lead to inter- esting results. (Gamespot) 2.4 Rigging a Character for Animation Usually, a 3D model by itself is a static mesh which cannot be animated effectively. Especially when working with the model of a creature, creating a system for mov- ing its limbs is important. The task of creating a system for moving a character by  7 creating bones and joints is referred to as a rigging while the structure itself can be referred to as a rig, armature or skeleton. (Slick 3, 2016) An armature is built of bones that form chains and are placed inside the character model. There are various kinds of bones and joints that can be used in rigging. The two main types of bones are the deforming bones and the control bones. De- forming bones are the bones that are directly connected to certain parts of the 3D mesh. Moving, rotating and scaling a deforming bone will also alter the character model accordingly. Meanwhile, control bones are used to control other bones, di- rectly or indirectly. Moving a control bone will not directly affect the surface of the model, but it may move other bones which again deform the mesh. (Blender 2.78 Manual) When it comes to creating a system to move the joints, the two main types are Forward Kinematics (FK) and Inverse Kinematics (IK). Forward Kinematics refer to a system where the bones are moved one by one to create a pose. For example, posing hands is usually done by first bending the bones closest to the palm and moving outwards. Inverse Kinematics are a system where the bone at the end of the chain is moved and the program moves the bones below it accordingly. For example, if used in a hand, the fingers could be pose by only moving and rotating the tip of the finger. IK systems are though somewhat tricky to create and can move in unpredictable ways if not properly executed, thus requiring the artist to do some manual cleanup later. IK systems work well when used in arms and legs while FK can be used for fingers and spines. (Slick 3, 2016) Once the armature is finished, the model is attached to it in a process called weight painting. Some programs like Blender have the option to automatically assign the character mesh to the bones but the automatic weights usually need to be ad- justed. A common example of a problem that can occur with automated weight painting is the character’s clothes or hair moving at a different pace than the body part it should cover when a limb is moved, thus causing the body to clip through and leave the clothing or hair behind. (Pluralsight 2015) Depending on the com- plexity of the rig and the model, the process of creating a working rig can take anything from a few hours to several days or even weeks. (Slick 3, 2016)  8 3 THE STRUCTURE OF A GOOD CHARACTER MODEL There are several things that can go wrong when creating a 3D character. The model’s topology can be a mess, the UV layout can be a pain to work with or the rig can be prone to breaking the model when animated. When a character is crafted poorly, it can be quite an eye-catching sight, if not a disturbing one, while a well-made one can greatly help the player focus on the game itself. It is good to know what the good practices of 3D modeling are before spending too much time trying to create a complex character for a game. 3.1 Polygon Count In the case of video game characters, topology and optimization are important. Because in games the player chooses what their character does, everything needs to be rendered in real time which means every image you see is created as you play. Creating these images takes processing power and time depending on how much detail the character models and environments have. If the models would be too complex, the player would have to wait to see the results of their actions be- cause rendering the images would take so much time. (Silverman, 2013) But what is a good amount of detail for a character model to have? It all depends on the platform the game is aimed for. The Unity User Manual states that the ideal polygon count for mobile games is 300 to 1500 polygons, while desktop platforms can handle character models of 1500 to 4000 polygons. If there are multiple char- acters on the screen at the same time, the polygon count may need to be reduced to lessen the load (Unity User Manual (5.5), 2017). Movies are not interactive like games which means they can be pre-rendered. This means rendering time is not an issue and the characters and environment can be as detailed as the creators wish. (Silverman, 2013) In video games, cut scenes can be pre-rendered and some games really do use cut scenes to show off their artists’ skills. For example, the Nintendo DS version of Final Fantasy IV includes an opening cinematic with character models so detailed the console would have  9 no hope of rendering them in real time. The game itself is played with heavily styl- ized and simplified character models. (Image 4) Image 4: Same game, same characters, different priorities. Screenshots from Final Fantasy IV’s Nintendo DS release. It is important to take into consideration the game’s needs when creating the char- acter models. In the right picture above, we can see the in-game character models’ hands are very simple to the point of resembling hooves more than human hands. The characters in the game simply do not need individual fingers so the modeler decided to leave them out. Image 5 shows two different kinds of hands a game character might have. Another point to consider is whether the character needs a three-dimensional mouth or if talking can be animated with a 2D texture. It all de- pends on how close the characters will be to the viewer and what the character will do. Small simplifications like these can quickly reduce the character model’s polygon count.  10 Image 5: A hand with individual fingers and a hand with 3 fingers fused together. (Ward 2011) There is though one thing where a character model with a high polygon count is often used in games and that is in the creation of texture maps. A certain texture map is used to create the illusion of detail that responds to light without giving the actual model these shapes with polygons. This map is called the normal map. A normal map uses three different colours to store the direction of the normal of a pixel. These maps are created by creating a low polygon and high polygon version of a character and using a 3D program to in a way project the high polygon model’s details onto the low polygon model. When the normal map is applied to the low polygon model, its surface will respond to light the same way as the high polygon model would while still remaining quick to render. (Polycount, 2016)  11 3.2 The Default Position When a character model sheet is made for a 2D production, the character is often drawn in a pose that shows how the character’s limbs are drawn in different posi- tions. (Masters 2015) In a 3D model sheet this is not advisable as the character’s pose will be defined by it. In 3D games, characters are usually modeled in a pose where the arms are extended straight to the sides. This is referred to as the T- pose. The point of the pose is to make the character easier to model, rig and ani- mate. After all, if the character’s arms were down and relaxed, it would make it difficult for the modeler to work on the character’s sides, armpits and the underside of their arms. Especially weight painting in the rigging phase can get messy if the arms are close to the torso. Still, the traditional T-pose has its problems as well. In T-pose, it is harder for the animation software to figure out which way the arm should bend and the stretching on the shoulders is more prominent when the char- acter lowers their arms. (Engländer, 2015) In newer games, you can see the characters’ arms at a slight downward angle. The change is quite recent and there haven’t been many official explanations for it. One theorized reason for this is the fact game characters rarely need to raise their arms straight above their head which means the midpoint of the usual range of motion is lower. Having the arms slightly down in the base pose makes the topology of the shoulders more natural and reduces stretching of the shoulders when the arms are being moved. (Polycount, 2011) 3.3 Clean topology When a model is brought into a game engine, it will be triangulated. This means every polygon will be converted into tris, triangle-shaped polygons. If the model already consists of tris or quads everything should be fine. After all, tris need not be triangulated further and quads can only be triangulated in two different ways. Though the artist should not mix tris and quads too much. They should stick to one of the two polygon types. Usually quads are a better choice as they can be trian- gulated for game engines or subdivided for sculpting as needed. (Slick, 2016)  12 Problems may though arise if there are polygons with more than four edges edges, so-called n-gons. The more edges a polygon has, the more different ways there are for the engine to triangulate it and the results are unpredictable. (Silverman, 2013) Image 6 shows all the ways a pentagon-shaped polygon could be triangu- lated. Image 6: Triangulating a pentagon. (Silverman, 2013) When creating a character model, it is a good idea to try to match the topology with natural muscle lines. This ensures the character’s body will behave more nat- urally when animated. Another point to remember is keeping the topology clean and even. The modeler should aim to have the body’s polygons form a neat grid (Ward, 2011). One thing that can cause problems in topology is having poles in the mesh. A pole is a vertex where 5 or more vertices connect. Poles can be used in certain areas of the model, as shown in image 7, but they generally can cause problems if the modeler does not know where to place them. (Polycount Del, 2011) Poles disturb the flow of edge loops and might even cause an edge loop that seemed fine to suddenly spiral around the model several times, making it difficult to later on add edge loops to the model.  13 (Image 7: Examples of where poles can be placed on a character’s face to help avoid triangles among quads. (Polycount Del, 2011) When creating any 3D model, the modeler should remember to remove any poly- gons that cannot be see. In a first-person shooter, they might want to delete the polygons on the underside of the player character’s gun. Similarly, if a character has their body and clothes modeled separately, any parts covered by clothes should be deleted from the nude model of the body. When it comes to edge loops, any loops that do not alter the model’s shape should also be removed (Ward, 2011). 3.4 The structure of the joints When it comes to characters that need to be animated, the structure of joints needs to be planned well. There are several ways to handle the structure of a 3D char- acter’s joints, some more natural-looking than others. When a 3D character is an- imated, its body parts will deform. Some polygons will stretch while others will shrink. Problems though will arise if the areas that move will cause the character’s  14 body structure to collapse. Image 8 shows 3 different joint types and how they look when bent to a 90-degree angle. The third type demonstrates a joint that collapses when bent. Image 8: Different kinds of joints. (Polycount, 2015) 3.5 Good practices When creating assets for a video game, there are certain things the artists should do to make each other’s work easier. After all, video game development is a team effort and all the work needs to be combined eventually. The artists should to- gether decide on things like world scale and naming conventions. There is also always the chance that someone must pick up someone else’s unfinished work  15 and finish it. If this happens, it is especially important that the 3D files are well organized and easy to navigate. 3.5.1 World Scale World scale means setting the measuring system and grid size of every artist’s 3D program to be accurate to the game world. Centimeters are a commonly-used system in most studios. Setting the grid size to be the same on every artist’s com- puter makes it easier to keep all the assets in scale to each other and helps the technical artists to create shaders that work in the chosen scale. Without a world scale system like this, the environment assets may turn out massive compared to the player character or the lighting may turn out too strong or weak. (Rinaldi) 3.5.2 Naming Conventions regardless of whether the artist is working on 3D models or 2D textures, naming everything is important. Naming the files is a given, but also every separate object or layer should be named in case the piece needs to be later on reworked either by the original artist or by someone else. If every object has generic names like “Sphere1” or “Cube3”, finding the object that is a specific piece of the character’s armour wastes time. Every object and material should have a name that is easy to recognize. This also goes for creating 2D textures; trying to find the layer that has the character’s shirt could take a while if every layer is named “Layer#”. Naming everything is especially important when rigging. Because bones rarely are too unique in appearance, every bone should have a name that is easy to recog- nize even when the character has been pulled into a pose that is hard to read. For example, a bone that controls the rotation of a quadrupedal character’s left front knee could be named “FrontKnee_L”. One commonly used naming convention is the CamelCase. In this naming con- vention, everything is written without spaces and with every new word capitalized.  16 In programming the first word is generally not capitalized. For example, “redBrick- Wall” could be a material named in CamelCase. Using the same naming style as the programmers makes it easier for them to work with the art assets too. (Rinaldi) 3.5.3 Version control There is always the chance that recent changes to a model need to be rolled back in one way or another. The modeler may have removed a part of the model without noticing or maybe they accidentally applied a modifier and are unable to trace back the changes it made. This is where version control proves useful. Whenever the modeler is comfortable with their progress, they can save their current work as a separate file, like save states in games. If the artist realizes they made a mistake and are unable to undo enough steps, they can always load an earlier version of the model and take a different path. (Van Gumster, 2016) The artist can use either version saving system that are already included in the program, Blender for example has the option to allow version saving, or they can use external programs. Different versions are usually numbered incrementally, for example, “mainCharacter_02”. Some people may also name different versions by adding the date it was saved on to the end of the file name.  17 4 THE BASICS OF BLENDER Blender is an open source 3D modeling program which can be downloaded for free at www.Blender.org. The program is free to use even for commercial projects and can be used for every aspect of 3D modeling from the actual modeling to sculpting, texturing and rigging. Blender is though quite different from most other 3D programs due to its high customizability and reliance on hotkeys for most tasks. This means the program can have a steep learning curve when someone with history with other programs decides to give it a try. This chapter will very briefly explain the very basics of Blender for those who are unfamiliar with its functions. 4.1 Setup Before beginning the actual 3D modeling process, there are a few things that one might want to change in the Blender user preferences window. The user prefer- ences can be accessed from the File-tab or by using the “Ctrl Alt U” shortcut. The first thing to change is the history states, or global undo. These are the amount of times the artist can undo their steps and by default this number is quite low. The global undo can be altered in the bottom-left of the Editing-tab of Blender User Preferences window. (Image 9)  18 Image 9: The Editing-tab of Blender User Preferences. The Global Undo slider is highlighted in red. For someone whose first experience with 3D modeling was 3DS Max, Blender’s default controls soon proved quite frustrating in how different they were. Luckily, it is possible to customize them in the User Preferences window. Whether one wishes to use Blender’s own controls or change them to match 3DS Max’s controls is their own choice. One should though remember, that if they decide to change any settings, they will need to make the same changes on every computer they work on from here on out. Writing down these guidelines somewhere where they can always find them easily is not a bad idea. First, by default Blender uses the right mouse key to select things instead of the left mouse key which is used in most other programs, including 3DS Max. This can be changed in the Input tab of the Blender User Preferences. Secondly, Blender uses different controls to rotate and move the 3D view. These can also be ac- cessed in the Input tab by selecting the following route in the drop-down menus on the right: 3D View  3D View (Global). In the 3D View (Global) menu, the settings to change are titled “Rotate View” and “Move View”. The key bindings can be changed by clicking the current key combination on the right side and pressing the desired keys on the keyboard. Image 10 highlights all the changes and shows  19 how the Input tab should look afterwards. The changes need to be saved by press- ing the “Save User Preferences” button in the bottom-left or else none of them will be applied. Image 10: The changes needed to make Blender controls more similar to 3DS Max. 4.2 The interface It is recommended to open Blender and examine it first-hand instead of only relying on the images of this guide. There are several tabs, menus, windows and buttons for a modeler to examine which makes showing everything in still images difficult and redundant. Blender is also a highly customizable program which can look very different depending on who is using it. When Blender is first opened, the screen will look something like what can be seen in image 11.  20 Image 11: The first look at the Blender interface. On the left side, there are several tabs that can be used to create and edit objects. The tabs will be different depending on what mode the window, or viewport, is currently in. When first opened, the first viewport is in Object mode, as seen from the small menu at the bottom of the window ( ). In Object mode, the most important tabs usually are the “Create” and “Tools” tabs. In the Create tab, new objects can be created while in the Tools tab these objects can be edited, duplicated and joined together. By clicking on the button with the Object Mode text, the mode can be switched to Edit mode. While simple objects can be altered in Object mode, Edit mode allows a wider range of actions than can be performed on the object. It is important to note that shapes created in Object mode will be separate objects with individual settings and names. Meanwhile, creating new shapes in Edit mode will add the shapes to an existing object and apply any settings and modifiers from that object to the new shape. For example, if on object has a Mirror modifier, cre- ating a new sphere on one side of the scene in Edit mode will create an identical sphere on the other side, while creating a new sphere in Object mode will only create one sphere wherever the modeler wishes.  21 In Edit mode, the Tools tab has a wide range of tools to edit the initial shape with. Some of the tools are very self-explanatory while others can sound alien to a be- ginning 3D artist. The Extrude tool can be used to extend a selected face or edge to a chosen direction, the Inset tool can be used to in a way copy the shape of the selected faces inside the selection, while the subdivide tool can be used to split edges and faces in half. The Loop Cut and Slide tool can be used to add edge loops to wherever the modeler wishes. While the Loop Cut and Slide tool is active, the amount of created edge loops can be adjusted with the mouse scroll. The various buttons around the Object Mode button ( ) at the bot- tom also depend on what mode the window is currently in. The three first buttons will always be there while the rest will change. The menu can be used to change the model’s appearance on the screen. By default the model will be shown as a gray shape, but here the model can be shown to be displayed as a wireframe, textured model or even a rendered image. It should though be noted that this ren- der is a preview and takes a while to load properly. Even slightly moving the view in Render view will cause the scene to be recalculated which will show as an eye- straining storm of pixel noise. The menu is where the modeler can select the current pivot point. By default, all changes will be done by the current selection’s median point, but here the pivot can be set to, for example, individual origins which allows several connected pol- ygons to be scaled individually rather than in relation to each other. To demon- strate this, let us look at a cube with three subdivisions. Subdivisions can be added with the “Subdivide” tool and the top faces can be extruded out with the “Extrude Individual” tool in the left toolbar. This will extrude the faces so they are separate towers even if they appear to be an even face. Scaling these faces by the median point or by individual origins will create very different results as can be seen in image 12.  22 Image 12: The difference between scaling a group of extruded surfaces by median point and by individual origins. The next buttons ( ) will rarely be used in this guide as most of their features can be used with the use of hotkeys, but they can be used to change the Manipulator’s (the Gizmo in 3DS Max) mode. The arrow lets the ma- nipulator to be used to move the object, the arc lets the object be moved and the club-looking shape can be used to scale the object. The last drop-down menu offers different choices for the orientation of the Manipulator. In Object mode, next to the Manipulator settings are two sets of squares ( ). These squares represent different layers where objects can be placed. A layer with any sort of content will have an orange ball inside it. Several layers can be selected at the same time by holding Shift while selecting them. In Edit mode, the layer boxes are replaced with the following buttons,. These represent different ways of selecting parts of the 3D object. The first one will allow individual vertices to be selected, the second one will select edges be- tween the vertices and the last one will select faces. A more peculiar option though is the button. This button will allow vertices, edges and faces to be seen and selected through the model. This button can be an invaluable tool when selecting areas with a lot of detail as it allows even faces hidden small crevices like a mouth or behind an ear be selected with ease.  23 Similarly, the tool can be a nuisance if it is left active when trying to focus on a specific area of a 3D model. The last set of buttons to look at is the group. These buttons represent the edge snap tools. When the magnet button is active, parts of the object can be moved in even increments. The menu next to the magnet where the constraints of the snap can be selected. For example, when the snap target is set to vertex, moving the selected vertices will always snap onto the vertex closest to wherever the selection is moved. Moving away from the buttons of the toolbar, in the top-right corner is the Outliner which shows the structure of the scene. It will list out all objects created in Object mode ranging from cameras to actual shapes. The names of each object can be changed by double-clicking the name. Pressing the button next to the name of an object will open that object in Edit mode. This is a quick way to switch between objects without needing to visit the Object mode in between. The eye symbol will hide and unhide the object, the arrow symbol will toggle whether the object can be selected and the camera symbol determines whether or not the object can be seen when rendering the scene. If an object suddenly disappears without reason in ren- dering, having accidentally clicked the camera symbol can be the reason. The list of tabs under the Outliner is the Properties panel. Here the settings of a single object can be altered. Explaining all the tabs would take a very long time but luckily only two of them need to be explained for now. The tab allows the addition of modifiers. When creating a 3D character, some notable modifiers are the Mirror modifier which allows the easy creation of symmetrical creatures and the Solidify modifier which can be used to give even thickness to pieces like clothes and armour. Meanwhile, the tab is where materials can be created and assigned to an object. In the Material tab, the objects colour and texture can be changed as needed. As mentioned, Blender’s interface is very customizable. All of the windows ex- plained can be changed into any mode from the small drop-down button at the left side of each window. The Outliner can be made into a 3D viewport, the Timeline can be made into a 3D viewport, even the Info tab at the top can be turned into a  24 3D viewport. The original 3D viewport can also be split into ten more viewports by pulling on the top-right corner ( ). The options are nearly limitless, if not actually limitless. Extra viewports can be merged together by grabbing the corner again and pulling it towards a window that is the same width or height as the current one. Image 13 shows all these described viewports in action. Image 13: Blender’s interface split into a plethora of 3D Viewports. As a last notion, the + symbol at the top-right of some windows can be used to open an extra settings menu. Depending on the type of window currently open, the contents of this menu wary. 4.3 Hotkeys Blender is a program with a massive number of hotkeys. Many of the actions can be found as buttons, but most actions do not have a button. For example, creating a face between two edges or merging selected vertices are actions that do not have a button on screen. An alternate way to finding these actions is searching for it by pressing the space bar and writing the name of the command. Learning all the hotkeys by heart may seem like a daunting task at first, but the most used  25 commands will quickly become a second nature. Having a list of the hotkeys though can be useful, especially if taking a break from using Blender. All the hotkeys used in the creation of the character model later in this thesis will be explained in the tutorial itself. Attachment 1 has a list and short description of all the hotkeys used in this project along with a few other hotkeys that could be useful elsewhere. 4.4 Modifiers Modifiers are tools similar to filters in photography. Modifiers can be added and removed from objects as needed and they can be used to add different effects to the piece. When modeling characters, one of the most useful modifiers is the Mir- ror modifier. The Mirror modifier will copy all the topology on one side of the ob- ject’s origin point along whichever axes the modeler wishes. The Mirror modifier makes it easy to create symmetrical characters and speeds up the process greatly. Another useful modifier is the Solidify modifier. This modifier works best when cre- ating objects like armour or clothing. The Solidify modifier will add thickness to any 2-dimensional plane and give it thickness. For example, the modeler can first cre- ate a long coat for the character out of flat planes and later use the Solidify modifier to make the jacket thicker. The Solidify modifier also ensures the piece is the same thickness everywhere. (Image 14)  26 Image 14: How the Solidify modifier gives thickness to a flat object. Some characters may have parts that need to be copied and pasted in an even line. For example, a fluffy tail could be created out of cones that are copied inside one another or the ridges of a dragon could be copied in a line along the creature’s back. For such pieces, the Array modifier can prove useful. The Array modifier will copy the object as many times as needed into a set direction. It is also possible to have the Array follow a set shape like a curve. There are several other modifiers, but their effects may be less useful when cre- ating a simple game character. The Multiresolution and Subdivision Surface mod- ifiers can be useful when sculpting a very detailed 3D character but they hardly are beginner-level features.  27 5 3D CHARACTER CREATION STEP BY STEP This chapter will show the process of 3D character creation step by step. The first part shows how to draw the character’s 3D model sheet and gives tips for keeping all the views consistent. The latter parts will focus on different parts of creating and preparing the 3D character model for use in video games. The character featured as an example is not going to be used in a game but will be created in such a way that it could be used in a game like The Elder Scrolls V: Skyrim. The character will have all pieces of clothing modeled separately so they can be equipped and un-equipped as necessary. Overlapping pieces of clothing will have some polygons removed to make the model lighter. If it cannot be seen, it will not be rendered. The aim is to have the character’s body consist of about 5,000 polygons while wearing any combination of clothing while the character’s head should have about 1,000 polygons. This way the whole character model should have about 6,000 polygons. The example character will be referred to by his name, Aroleir, through the course of this guide. 5.1 Creating the 3D model sheet Before the artist begins creating a 3D model sheet, they should already know what kind of character they are creating. This goes both for the character’s clothes and the level of realism the model would have. The T-pose used for a 3D model sheet strips the character of their personality which is why designing a character in this stage is not ideal. Polishing the character’s design before creating the 3D model sheet also ensures that the artist will not need to share their focus between creat- ing an interesting design and keeping all angles of the model sheet consistent. One can also use a pre-made model sheet but the artist should remember to make sure they have permission to do so if they plan on publishing the model. It is also good to note that a pre-made 3D model sheet may have the exact same mistakes  28 that were mentioned in the earlier chapters and thus using them to make a 3D character might be more difficult than anticipated. 5.1.1 The front and back views The first stage of creating a 3D model sheet is drawing the nude character’s front view without anything extra. This makes figuring out anatomy easier. Even if the character would never be shown without their clothes in the game, the nude refer- ences make figuring out topology and joints easier in the 3D modeling phase. The front view is a good starting point as the back view can later be traced over it and the side views can use the front and back views as points of reference. It is up to the artist to decide whether they wish to have an unclothed version of the character from all directions or not. Front and side views can be enough refer- ence for creating the naked character model for a humanoid character. Aroleir’s front and side references were first drawn nude, but the back view was drawn only after all the clothes of the front view were done as time was limited. If the character is meant to be heavily stylized, the artist can make their own deci- sions on the character’s proportions. Meanwhile when creating a realistic charac- ter, finding reference material is more crucial. One should remember that in full body photos you rarely see an orthographic projection of the person. Especially feet often are shown slightly from above in drawings and photos of people. In a 3D model sheet, the feet should though be drawn directly from the front as shown in image 15. Image 15: Orthographic view of feet and shoes.  29 The character’s pose in a 3D model sheet is very neutral and there are several graphs that show the proportions of a human body. The first step of drawing the character would be figuring out their height in comparison to their head. This height varies greatly depending on the style and species of the character, but in Aroleir’s case the character’s height was set to be 7½ times the size of his head. Anything less than this would give the character a cartoon-like feeling while anything taller might take him towards an idealized style like in superhero comics. Image 16 shows a few examples of men with different body proportions. Image 16: Different ways to draw a man. (Andrew Loomis) A scale for drawing the character can be created quickly by drawing an egg shape that represents the character’s head and copying it as many times as needed. This  30 scale can then be used to block out the basic shape of the body per the artist’s vision for the character. Aroleir is meant to be fit and toned, but he does not need the hourglass figure and pronounced muscles of a superhero. Image 17 shows the very rough sketch of his proportions alongside the finished lines. Image 17: Body sketch and finished lines. The artist can and should speed up the drawing process by first drawing only half of the character and mirroring it if the character has parts that are symmetrical. After all, the 3D artist will most likely also use mirroring tools to speed up their work if the character’s design allows it. If the references would not be mirrored like this, it would be highly likely that the halves would not be perfectly aligned. This would be visible in the 3D modeling phase, as one of the reference’s halves would be slightly off. While this would not be much of a problem, it would still be an annoy- ance that could easily be avoided. Another point to this is the fact that perfectly  31 symmetrical faces can easily look strange. If the references are mirrored, the artist can in that phase already try to make the face’s proportions look pleasant despite the perfectly symmetrical features and thus make the 3D artist’s work a bit easier. When the clothes are being drawn, it is a good idea to place them on a separate layer so the nude version can also be given to the 3D artist. The nude version helps the modeler tell where the character’s joints should be even if their clothes would be loose-fitting. It also usually is easier to add the clothes after the base topology of the model is done. If the game has several characters, the nude char- acter model can also be edited to make different characters quickly. In Aroleir’s case, every piece of clothing was drawn on a separate layer in the front and side views due to the way he should be able to wear different clothing combinations. The back view though was drawn fully clothed as the side view gives enough in- formation about how his clothes work on the back. Attachment 2 shows the fin- ished reference sheet with different combinations of clothing. Once the front view of the character is finished, with or without clothes, creating the backside is a straight-forward process. First, the line art of the front view is copied and mirrored. By deleting all the lines within the outline, we get a base to draw the back view around. This way the silhouette remains the same in both front and back views. Certain details that continue from the front to the back, like edges of clothes, belts and such, can be marked on the back view and Image 18 shows the stages of this method.  32 Image 18: Creating the back view with the help of the front view. 5.1.2 The sides and other views After the front and back views are done, creating the side view is just a matter of combining the information of the two references. The front and back views are to be placed on opposite sides of the canvas with enough space in between for the side view. To keep the side view consistent with the other two, straight horizontal  33 lines are to be used to connect matching parts in the front and back views. In certain drawing programs, like Photoshop, these drawn lines can be created by using the built-in ruler tools. If the used program does not have rulers, another way to do this is to use the rectangle selection tool to mark the area where a certain part is to fit. It is also possible to draw straight horizontal lines on a separate layer. Regardless of which method is used, it is easier to make sure every detail lines up with the other views when using these guidelines. Image 19 shows an example of how the rectangle selection tool was used in drawing Aroleir’s side reference. Image 19: The use of rectangle selection tool in positioning the studs of his armour. When creating the side view, it is a good idea to either draw the arm separately or leave it out completely. Depending on whether the character’s arms are extended straight to the sides or at a slight angle, drawing them on the side view is either useless or harmful. If the arms are at a slight downward angle, the arm will hide a part of the character’s sides. The downward angle in tandem with the orthographic view also makes the hand’s size and the arm’s proportions quite tricky to figure out. This is because the artist cannot simply use the character’s head’s height to measure the arm’s length as they could in the front view. The artist could though draw the arm straight from the side and the 3D artist could model the arm sepa- rately and attach it to the body when the limb is ready. In Aroleir’s case, the arm was not drawn from the side at all as the front and back views showed the details of his clothes well enough. One of the greatest challenges of creating the side view is staying true to the or- thographic view. If the character has for example a belt that hangs loosely at an angle, it can be difficult to figure out how it would look from the side. This is where using the horizontal guidelines is especially useful even if tedious. In Aroleir’s  34 case, this problem was apparent when creating the cut pieces of fur around his waist. From the front and back the shape was nothing special, but in the side view it took a few tries to find a shape that looked good. Image 20 shows a few exam- ples of how the furs could have been drawn while keeping the reference true to the other images. Image 20: Different shapes that are all plausible. The middle one is the final shape. 5.1.3 Bringing the references to Blender Before the character references can be brought to Blender, they need to be saved on separate files. All the references should be placed in the middle of the file and aligned to each other to make bringing them to Blender as easy as possible. This can be done by first placing all the references on top of each other in a 1:1 square shaped file, saving a copy of it for each view and deleting the extra layers from each file. This way there will be a separate file for each view while also having all the views be on the same spot on the square file. Once the references are saved as PNG files, they can be brought to Blender by using the Background Image feature. On the very bottom of the right sidebar of  35 the 3D Viewport there is a tab titled “Background images”. By ticking the box and clicking the “Add image” button, you can select a reference picture from your com- puter. After the image has been imported, by default it will show in all main views (Front, back, left, right, top, bottom). The images are automatically hidden when the character is rotated so they will not clutter the viewport. It should be noted that the background images only show in orthographic view (Numpad 5) as in perspec- tive view they would be inaccurate. A single view can be selected in the drop-down menu titled “Axis” (Image 21, point 1). By default, the images are displayed behind the model. This means they cannot be seen through the model which defeats the purpose of the reference images. To see the images through the model, the images should be set to be transparent (Image 21, point 2.) and displayed on the front of the viewport rather than the back (Image 21, point 3). This way the model can be seen through the reference im- ages. The images are placed in the middle of the scene so the base grid crosses over the character. This can make creating the character more difficult as then the cen- ter of the grid cannot be used as a point of reference when positioning different parts of the character’s body. The reference images can be moved higher in the scene by adjusting the Y value near the bottom of the background image settings (Image 21, point 4.). The reference should be placed so that the character’s feet rest on the grid. This way the character’s origin point will be between their feet which makes adding him to the game easier. The value slider changes in incre- ments of 0.1 which can be too inaccurate. The value can be set manually in more detail by clicking on the number. In Aroleir’s case, his side reference was only drawn of the left side but it would make matters easier to be able to see the reference from both sides. To do this, the same image was applied twice and was set to be displayed on both left and right side. In the right side, though, the reference had to be mirrored so he would face the same way as on the left side. (Image 21, point 5.)  36 Image 21: The settings for the front background image. Here the settings are shown as they were when making the model for Aroleir. 5.2 Creating the 3D Model The modeling process begins from the character’s torso as it gives the base for the model’s level of detail and overall structure. Creating a cube is a good start as it is easy to add edge loops to such a simple shape. The cube is resized so it is about the size of the character’s chest and in Edit-mode it is moved to the height where the character’s chest should be. The move is done in Edit-mode because if the object would be moved in Object-mode, the model’s origin point would move with it. In Edit-mode, the object’s origin remains at the middle of scene. If the origin point would move elsewhere, it would later cause problems in the use of mirroring tools and in bringing the character to a game. If the origin has moved from its place by accident, it can be returned back to its intended place by first moving the 3D Cursor to the middle of the scene (Shift + S  Cursor to Center) and then moving the Origin to the 3D Cursor (Shift + Ctrl + Alt + C  Origin to 3D Cursor). The origin can only be moved in Object-mode.  37 Before beginning to edit the cube, the “Limit Selection to Visible” ( ) setting should be turned off. The button can be found in the bottom toolbar in Edit-mode. When this button is turned off, vertices can be seen and selected without the sur- face of the model obscuring the view. Basically, this shows the model’s full wireframe through the surface polygons regardless of what angle the model is viewed from. The first change to the cube is setting it up for the Mirror modifier. A vertical edge loop is added to the middle of the cube with the “Loop Cut and Slide” tool. This edge loop marks the character’s middle line. Next, all the polygons on one side of the middle line are deleted, leaving one half of the cube. This is when a Mirror modifier can be added. This modifier copies all changes done to the model in real time, which means only one half of the character needs to be modeled by hand. The Axis of the modifier should be set to X which is the default setting. (Image 22) Image 22: The final settings of the Mirror modifier. The Clipping option is disabled by default but it should be enabled most of the time when creating the character. The clipping option makes sure the middle line cre- ated before stays where it should and prevents vertices from moving to the other side. If vertices would end up on the wrong side, the mirror modifier would copy  38 them twice and create overlapping faces. (Image 23) The Clipping option can oc- casionally be disabled for a moment if the character model requires separating vertices on the middle line. Image 23: An example of troublesome topology that may occur if the Clipping set- ting is turned off when modeling. The highest vertex has slipped to the wrong side and thus created overlapping faces and the clean middle line is lost. 5.2.1 The torso After the Mirror modifier has been prepared, the process of turning the cube to the character’s body begins. The Box Selection tool (B) is used to select and move the cube’s top vertices where the character’s shoulders will be and the lower ver- tices are taken to where the character’s hips will be. This is done from both the front and the side views. Next, horizontal edge loops are added to the shape and the created vertices are scaled so they line up with the character’s waist’s silhou- ette in the reference images. Once the horizontal edge loops are in good places, vertical edge loops are added to both the model’s front and sides. Image 24 shows the added edge loops without any changes to the model’s silhouette to show how  39 many edge loops were added on each side. When crating the actual model, it is advised to not add these loops all at once as they clutter up the views fast. Image 24: In the beginning, four horizontal edge loops were added. After them, two edge loops were added to the Y axis (front-back) and three were added to the X axis (right-left). More edge loops can be added as the need arises. This phase is mostly about moving vertices around and trying to create the silhou- ette for the character’s body from the front and the side. The views can be changed by pressing the Numpad keys 1 and 3. The aim is to find balance between soften- ing the character’s features and being mindful of the game engine’s limitations. This job can be made easier by selecting several vertices and using the scaling tool (S) to round out the shapes, and by having the Edge Slide tool move vertices along the existing edges by pressing the G-key twice after selecting the vertices. Once the shapes have been matched to the front and side references, the model seems quite competent already. However, when the view is rotated to any other view, the model turns out to have right angles instead of rounded shapes. Before rounding the corners, this is a good place to create holes for the character’s arms  40 and neck. The main thing to remember here is that the arm holes should have an even number of vertices and the removed polygons should be chosen so the ver- tices around the hole are easy to move into a round formation. In Aroleir’s case, four polygons were removed from the side (arms) and the top (neck). The arm and neck holes were then rounded to match the references. At this point the holes can be very rough. Image 25 and 26 show two ways of creating the arm hole. The first one shows the arm hole on Aroleir, while the second one features an earlier char- acter project. Image 25: The creation of the arm hole on Aroleir. The faces marked with blue were also deleted to create the hole for his neck.  41 Image 26: This character was much lower poly than Aroleir and even a peculiar solution like this worked well. When it comes to rounding out the character’s sides, there should be one edge loop on both the front and the side that won’t be touched at all. This will be the loop that determines the character’s silhouette as shaped before. When creating a realistic character, these silhouette-determining loops are usually the middle loop or the one behind it, while on the front and back views this may be the loop next to the middle line. After all, the human body usually has a slight dip between the shoulder blades and chest muscles. On a more cartoony character, the silhou- ette-determining loops may be the very middle loops as a small detail like the dip between shoulder blades is not necessary. Once the silhouette-determining loops have been selected, all the other edge loops will be moved inwards to get rid of any angles that stick out. The first edge to change should be the very corner of the shape. All the vertices of that one edge should be selected and moved along the Y and X axis so they would make a smooth curve with the two edge loops next to them. Next, one edge on both sides of the currently selected edge are selected and moved again to make a smooth  42 curve. This will be repeated until every vertex except the silhouette-determining ones have been moved and the shapes have been smoothed out. This goes also for the top vertices that would become the character’s shoulders. (Image 27) Image 27: Moving the vertices of the corner to smooth out the shape. The silhou- ette-determining edges that were not moved in this phase are marked with blue. When creating the front of the torso of a more realistic character, the loops can be edited so they follow natural muscle lines of the body. Namely, this means curving the top loops so they form the shape of chest muscles. If done right, this will ensure the chest muscles will move naturally when the character moves their arms and the shape of the chest will be obvious even without textures. (Image 28)  43 Image 28: The finished torso. It should be noted that one of the edge loops around the rib cage was removed later in the project as it was not needed after all. Because a character’s arms are usually down and in movement the shoulders are under a lot of stress, adding extra polygons to the shoulders will prove useful. The shoulders can be given a more natural shape by selecting the second highest edges of the arm hole and creating faces between them with the Face tool (F). The created face will be split into four with the “Loop Cut and Slide” tool. This will leave behind a triangle-shaped hole with 6 vertices. The edges of this hole are  44 selected and a face is created in the hole’s place again with the Face tool. The new polygon will then be split into two quads by choosing the middle vertices and creating an edge between them with the Join tool (J). Image 29 demonstrates the steps of this action. Image 29: Adding more polygons for the shoulders. In the last image, an extra set of polygons was added as it lined up with the muscles on the reference images better. Once the new polygons have been created, their vertices need to be moved so they match the references. The vertices of the middle line are selected and moved along the X axis to match the character’s silhouette from the front. After that, the vertices next to them are moved so they form a smooth, round shape. The vertices on both side of the middle line can then be moved around at the same time and scaled together to get the curve even. If needed, more rows of polygons and edge loops can be added to the shoulder. (Image 30)  45 Image 30: Shaping the shoulders. The green lines show an extra row of polygons that was added after shaping the top of the shoulder. Red lines show an extra edge loop that was added to help make the shape smoother. Next, the character’s lower body is prepared the same way the shoulders were. An equal amount of edges are selected from the character’s front and back and faces are created between them with the Face tool (F). The amount of edges cho- sen depends on how detailed the character’s crotch needs to be. In Aroleir’s case, one edge was selected. This edge was later on split into two but at this point just one edge was easier to work with. The face created between the edges are then Subdivided (S) into two and the new edge is pulled down to the level of the char- acter’s crotch. Image 31 shows the steps of this process.  46 Image 31: The beginnings of creating the character’s crotch. The lower body at this time resembles more a spike than a crotch. The created edges need to be subdivided further and the new vertices to be moved into a more rounded shape. It should be remembered that this middle line will be the dip be- tween the character’s buttocks, not the highest point of them. For Aroleir, two edge loops were added to both the front and the back of the crotch strip. (Image 32)  47 Image 32: Refining the crotch’s silhouette. Red lines and dots represent newly added edges while the pink ones represent the edge added in the last step. In the next phase the modeler needs to decide how many vertices they want the character’s legs to have. Like with the arms, the legs should have an even number of vertices and thus the leg holes here should be the same. At this point Aroleir’s leg holes had 14 vertices which I found to be a bit too high. The number of vertices can be lessened without removing any edge loops by extending down the charac- ter’s torso like demonstrated in image 33. The two highest edges are joined to- gether with a face. The face is then subdivided by the number of vertices that the character’s side has and the created vertices are joined to the side with the “Alt + M” hotkey. The silhouette will then be adjusted to match the references. This method lowers the amount of vertices in the character’s leg hole by two every time a new row of polygons is created.  48 Image 33: Lowering the number of vertices in the leg holes. In Aroleir’s case, it was enough to do this process only once. If the number of vertices still feels too large but the modeler doesn’t want to bring the character’s hip any lower, the modeler can add these extra faces to the crotch area. This though means the modeler will need to re-adjust the silhouette to match the refer- ences. Once the number of vertices feels manageable and the leg holes have been rounded to feel natural, it is time to begin creating limbs for the character. Image 34 shows how Aroleir’s finished torso looked like at this point.  49 Image 34: The finished torso. 5.2.2 The limbs and joints A character’s arms and legs are created separately from the torso. They will be made as new mesh in the same object. While in Edit mode, a new mesh is created. Cylinder is a good starting shape for limbs as the amount of edges can be set from the start. As Aroleir’s leg hole had 12 vertices at this point, a 12-sided cylinder was created. This cylinder needs to be re-sized so it is about the thickness and length of the character’s leg and then moved to where the character’s legs will be. The cylinder can also be rotated to match the reference images. (Image 35)  50 Image 35: The cylinder that is the base of the character’s legs. When created, the cylinder is placed wherever the 3D Cursor is. If the cursor is in the center of the scene, the cylinder will be placed so a part of it overlaps the middle line of the Mirror modifier. In this case, the modifier’s Clipping option needs to be temporarily disabled so the cylinder can be moved to its intended spot. Oth- erwise the Mirror modifier will prevent the cylinder’s vertices from going past the middle line and stretches the shape sideways. Once the cylinder is in its place, it is time to begin shaping it. First of all, the cap polygons on the top and bottom of the cylinder need to be removed. After this, the vertices on the edges of the hole are selected and scaled to match the references. The bottom of the cylinder should match the character’s ankles and the tom should be roughly the size of the leg hole on the torso. The edge will likely need to be rotated slightly to make connecting the cylinder to the torso easier. In side view, the cylinder should be placed so the ankles are in the right spot. The modeler needs to decide which vertices on the cylinder and the torso will be merged together. When the ankle is in the right spot, the modeler will pick which- ever two vertices are the closest to being aligned. These vertices are merged to- gether with the “Alt + M” command. The vertex on the cylinder should be merged  51 into the one on the torso so that the vertex on the torso does not move. This can be done by first selecting the vertex on the cylinder, selecting the one on the torso second and then picking the “At Last” command in the “Alt + M” menu. All the vertices on the cylinder should be merged to their corresponding vertex on the torso. (Image 36) Image 36: Merging the leg cylinder to the torso. Shaping the character’s buttocks can be quite tricky as the polygons easily distort when creating such rounded shapes. In this project, the challenge was ap- proached by creating edge loops to the top of the leg cylinder and shaping them according to the reference images one at a time. It may be helpful to enable smooth shading in this phase as it makes any sharp edges and distorted polygons more obvious. Smooth shading can be enabled by selecting all polygons of the model (A), opening the “Shading / UVs” tab from the left side of the 3D Viewport and clicking on the “Smooth” button under the “Faces:” option. Image 37 shows different phases of shaping Aroleir’s buttocks.  52 Image 37: Matching the upper leg to the references. The last picture shows how the model looks with smooth shading enabled. It is up to the modeler to decide what kind of joints the character will have. Aroleir’s knees and elbows got joints based on one of the examples on Polycount’s Limb Topology page (Polycount, 2015). The first step was to make a horizontal edge loops where the knee needed to be. Because the edge loops from creating the character’s buttocks are at an angle, the loops created under them will also be at an angle. They can quickly be rotated to be horizontal by selecting the vertices of the created loop and scaling them to 0 by the Z-axis (S + Z + 0). This needs to be done both to the knee loop and the ankle loop at the bottom. (Image 38)  53 Image 38: Scaling an edge loop to 0 by the Z-axis creates a horizontal edge loop. This image also shows an extra edge loop added to the thigh. This edge loop was used to make the character’s things look more rounded. To make the joint itself, a second edge loop is created under the one made before. About half of the vertices of these edge loops will be merged together one by one behind the knee, leaving two triangles on the sides of the knee. To make the joint so it will distort less when animated, another edge loop is added below the knee. The polygons on the front will limit the area that will stretch during animation while the two loops behind the knee will prevent the joint from collapsing. The steps described can be seen in image 39.  54 Image 39: Creating a knee joint. After the joint is done, the rest is merely shaping the leg to match the references. Since Aroleir is quite realistic, his legs were given four extra edge loops to reach a natural shape. The loops were though added one by one, starting from near the knee so all changes done to the shape would also work as a point of reference to the future loops. The vertices on the inner side of his leg were also moved a bit inwards to create a shape that mimics the shape of a muscular leg. Image 40 shows how the edge loops were added and edited.  55 Image 40: Shaping the leg. Note how in the last shot the vertices have moved to make the leg less tube shaped and more like the references. Creating the arms is mostly the exact same process as creating the legs; create a cylinder with as many sides as there are vertices on the arm hole of the torso,  56 resize and rotate it to match the references, delete the caps on both ends of the cylinder, attach the top vertices to the torso’s vertices, shape the arm to match the reference. One major difference is that the new edge loops cannot be simply scaled to 0 on Z-axis to create a straight horizontal loop. This means the joint should be made before connecting the arm to the torso. The joint will also be facing the opposite way to the knee since human arms bend to the opposite direction compared to the knees. Image 41 shows some shots from the creation of Aroleir’s arm. Image 41: The stages of creating an arm. 5.2.3 The feet and hands Creating the feet begins by creating a cube in Edit mode and moving it to where the foot should be. The cube is then adjusted to be roughly the shape of the foot from both the side and the front. Once that is done, an extra edge loop is added about where the ankle would begin. The top vertices of this new edge loop are  57 selected and moved to be about the same height as the back vertices of the foot. Image 42 shows these steps. Image 42: Blocking out the shape of the foot. Like with any other part that needs to be connected to the body, also the feet need to have a hole with the same number of vertices as the ankle holes on the legs. Aroleir’s ankle holes had 12 vertices which is quite easily matched on the foot by adding four new edge loops that cross over the area that will be connected to the body. The polygons on top should be removed to create the hole for the ankle but it is better to shape the foot before connecting the parts. The edge loops added on Aroleir’s foot can be seen in Image 43. Image 43: The edge loops are added to create an even grid. The orange and blue lines highlight the new edge loops. The right picture shows which polygons are deleted to make the ankle hole.  58 The feet would be covered by shoes most of the time so most details would come from textures. This means the feet don’t need refined details like separate toes. To make the foot into a shape that would look natural when textured, more edge loops need to be added to the sides that do not yet have many loops. The vertices of these edge loops are moved so they form more rounded shapes for the foot. Some points that are easy to miss are rounding the heel by pulling the bottom vertices inwards, raising the bottom of the foot on the inner side, and shaping the front of the foot to be the right shape when looked at from above. Image 44 shows what edge loops were added and the shape they were moved into. Note how the top of the foot’s shape tilts to the side instead of being a perfect half-circle. Image 44: Giving the foot its natural shape. The new edge loops are highlighted in red in the first picture. Once the foot is ready, it is attached to the ankle the same way as the other parts before. Matching vertices are selected on the foot and the ankle and then they are merged together with the “Alt + M” command. Some vertices may need to be ad- justed on the foot after being connected if the edges appear to tilt to one direction. After all, the foot was modeled from a cube and the leg from a cylinder, meaning the positions of the vertices may be different. Image 45 shows how the vertices of Aroleir’s foot were twisted and how they looked after being straightened. The “G + G” command for Edge Slide is useful when straightening the edge loops.  59 Image 45: Attaching the foot and straightening the twisted edge loops. 5.2.4 The hands The level of detail needed for hands depends on the game. Since Aroleir is meant to be like the characters in Skyrim, he should have detailed hands as his hands would be seen up close when in first-person view. All fingers are separate and modeled in such a way that they can be fully animated. The hand models are even detailed in third-person view as the characters move their fingers when casting spells. Image 46 shows an example of a player character’s hand in Skyrim.  60 Image 46: A character’s hand in Skyrim when the character is sneaking without having a weapon drawn. Image rotated for a more natural view. To create a hand, a cube is first created, resized and moved to the hand’s place. The corners of the cube are moved to line up with the palm as seen on the refer- ence images. The cube also needs to be scaled along the Y-axis to make the hand flat. Some edge loops need to be added to the shape and the modeler needs to keep in mind the number of vertices that the wrist hole has. Like before, these numbers need to match so the hand can be attached to the wrist later. Aroleir’s wrist has 8 vertices, meaning the hand got two extra edge loops on both sides of the future wrist hole. The vertices of the hand are moved to align with the vertices of the wrist hole so the hand will be easy to attach once it is finished. The polygons at the end of the hand piece are also removed at this point to create a hole for the wrist. Image 47 shows how the vertices that would be connected to the wrist were before and after adjusting to the wrist hole’s vertices.  61 Image 47: The beginnings of creating a hand. In the last picture, the vertices of the wrist on both the arm and the hand have been moved to align for easy merging later. The thumb is created next. The process begins by adding an edge loop to mark the part where the thumb will come. Some polygons on the side of the hand are selected and extruded (E) to create the root of the thumb. It is recommended to add an edge loop to run across the thumb sideways as it makes it easier to create the part of the thumb that is on the palm. The vertices of the extruded polygons are moved to match the reference pictures, only to have the polygons on top be extruded again. The finger will be extruded bit by bit until the desired shape is achieved both from the front and the side. The extrusions will give the finger a flat and angular form which needs to be rounded as the finger is being shaped. (Image 48)  62 Image 48: The stages of creating the base shape of a thumb. The thumb is not quite finished yet, though. The tip of the finger still needs to be rounded and a joint needs to be added. The shape of the fingertip depends on the level of detail and number of vertices at the modeler’s disposal. Aroleir’s thumb has eight vertices going around the finger which made it quite straight-forward to round up the fingertip. First, the finger’s overall shape needs to be adjusted. A thumb is not perfectly round but rather of a more oval shape. It is important to note that on a real human hand the thumbnail faces towards the back of the hand when all fingers are straight. While it may be tempting to make the thumbnail face up- wards along the Z-axis, having the thumb in such a position would make animating the hand more difficult later as the thumb would always rotate unnaturally. Once the shape of the finger has been adjusted to be more oval, the faces that cap the fingertip are deleted. To remake the fingertip, the long sides of the oval are connected with faces by selecting two opposing edges and using the Face tool (F). This leaves two triangle shaped holes on the sides of the new faces. The  63 holes are filled by selecting the edges and again using the Face tool. All the cre- ated faces are then split with an edge. This will leave behind two triangles on both sides of the thumb but the extra edges are needed to make the thumb tip. more rounded. The vertices created on this new edge are moved to shape the thumb tip. Image 49 illustrates these steps. Image 49: Creating the tip of a thumb. Note the thumb’s oval shape at the tip. The joints of the fingers will be quite different from the joints of the knees and elbows. On the fingers the joints will only have two extra edge loops running around the finger. On an eight-edged shape like Aroleir’s fingers, the joint will be symmetrical. The inner edge of the finger, or the edge towards which the finger will bend, will have its edges merged together like was done with knees and el- bows. This means the two parallel edge loops will be merged together by selecting corresponding vertices on both edge loops and merging them with “Alt + M” and  64 selecting “At Center”. This will create two triangles on both sides of the merged edges. The next step is to select three edges on the side opposite to the newly-merged edges. These three edges are subdivided with the “Subdivide” button on the left toolbar of the 3D Viewport. This creates a pentagon shape on both sides of the new edges. The pentagon needs to be divided into three triangles by connecting the ends of the new edge to the two bottom corners of the pentagon. All these steps are shown in Image 50. Image 50: The joints will be the same on all fingers. For extruding the other fingers there are no proper work-in-progress shots as the recording was cut off, but the steps will still be illustrated as screenshot edits. To create the fingers, more edge loops will be needed. These loops will later have their ends merged together so they will not interfere with attaching the hand to the wrist. These edge loops will be added to where the fingers will separate. Once the edge loops are in place, the fingers are created by selecting the faces the fingers should begin from and extruding them (E) one finger at the time. To ensure all fingers will be straight, these extrusions should be made while viewing the char- acter directly from the front or the back so the references are showing. Image 51  65 shows where the loops are placed and which polygons should be paired up for extrusion. Image 51: Extruding the fingers. The fingers need to be extruded one by one so they will not be fused together. After extruding the fingers, edge loops are added to each finger so they all have eight vertices around the fingertip. This way the fingers can be created like the thumb was. The only differences are that these fingers will have two joints each and these fingers will be round rather than oval shaped. Image 52 shows the stages of finishing up the fingers.  66 Image 52: Creating the fingers, fingertips and joints. The two last pictures show the finished finger joints from both sides. Creating the palm and back of the hand for Aroleir was an experimental process. The hands needed to have certain minute details like knuckles and be manageable to animate while also not having too many polygons. An added challenge was finding a way to merge together all the extra edge loops from the fingers before the edge loops reached the wrist. This meant the character’s hands would have triangles here and there and the true task was finding places where the triangles would not cause problems. The back of a human hand does not deform much when the fingers move which makes it a good place to hide triangles. Meanwhile, the palm deforms heavily when the hand moves but it also is easy to hide with the fingers, a weapon or spell effects. With these points in mind, the focus with Aroleir’s hand was on having the hand look good when in a neutral pose.  67 Both the back of the hand and he palm have small bumps at the root of the fingers. Triangles can be used to allow certain vertices to be raised higher than the ones next to them without the polygons bending in unfavourable ways. To create the knuckles, two edge loops were added to the hand. One was placed slightly under where the highest point of the knuckles would be and another was placed where the knuckles end. Image 53 show the actions done after this as they are difficult to explain in words alone. Image 53: The creation of knuckles and merging of the excess edge loops on the back of the hand. The triangles in the third picture were made by selecting two vertices and using the Join tool (J). Since the edge loops made when creating the knuckles also reach to the palm, no new edge loops need to be added. It may not be too important to have all the small shapes of the palm down perfectly, but they were still modeled on Aroleir’s hand as a personal challenge. Like with the knuckles, some triangles were created where the thicker areas at the root of the fingers were and the vertices were lifted and lowered to create the bumps. Certain vertices were fused together to rid of the  68 extra edge loops from the fingers. It is important to make sure the original edge loops that were aligned with the vertices of the wrist in the very beginning will remain in their place so the hand’s polygons will not twist when the hand is at- tached to the wrist. Image 54 shows some work-in-progress shots from the crea- tion of the palm along with the final product. Note how the edge loops added for the fingers were merged into the original edge loops of the hand. Image 54: Shaping the palm. One of the quads at the root of the thumb was split into two triangles so the shape would mimic the shape of a real thumb. Attaching the finished hand is done the same way as the other parts before. The modeler should check that all the vertices of the hand are aligned with the vertices of the wrist before merging them together or the wrist might end up with twisted polygons. 5.2.5 The head and face The way Aroleir’s face was created was based on a tutorial on 3DTotal.com (Roger, 2009). Certain things were done differently but the methods and order of creating things were similar. In this phase, it also turned out that Aroleir’s side and front references did not match up perfectly which made some areas harder to fig- ure out. In such cases the front view took priority over the side view. While the other body parts began with a simple shape like a cube or a cylinder, the face begins with a single plane. The plane is rotated 90 degrees around the X-  69 axis (R + X + 90) and scaled down so it is very small. While looking at the character from the front so the reference image is visible, the corners of the plane are moved so they form the first polygon of the corner of the character’s eye. Once the poly- gon is in a fitting place, the view is rotated so the character is seen from the side