top of page

Danny Sweeney and Das - 3D Artist Analysis

Danny Sweeney is a Senior Character Artist for Creative Assembly, and has created many characters for games over the years. He specialises in character art, and texturing surfaces such as cloth, armour and weapons. Having worked in the industry for seven years, he has a great understanding for anatomy and character design.



During a livestream, Sweeney talked about the techniques and pipelines that are used when creating characters. The pipeline for creating a character starts with rough art and concept art. This is done in the form of a Proxy, which is a very simple model of the character. This model focuses on the shape and silhouette of a character, and is used as a basic template of the character. This template can be used for multiple characters, as the riggers/animators can use it to build rigs in the models for the characters before they've all be created. This unfortunately means that a proxy model used by multiple characters cannot be changed. This is for multiple reasons, such as the proportions of UVs and the texture maps.



This proxy evolves into a modelled/sculpted character in high detail, which is then lowered to a lower poly count. This is done so that a character model isn't too big of a file to fit into a game, and having a higher poly version of the character saved means it can be added on later as a normal/bump map to give the model more detail.


Topology is also something that must be taken into account. Lower poly models means less polygons to work with, which can make animating simpler. In the livestream, Sweeney brought up different ways in which something called 'Limb Deformation' can be tackled. Limb Deformation would be seen when animating a bending knee or elbow, when the geometry warps as it moves. A way to tackle this is to add in edge-loops around where the limb bends. In Maya, I could do this by either using edge loops or creating a cut in the geometry with the multi-cut tool. However, since Sweeney seems to use software suited for sculpting, the equilavent of this tool in ZBrush are the slice brushes, which can cut into the existing geometry assuming there aren't too many subdivision levels on the model.


A tip Sweeney also offered was that triangle geometry is good to stop models from deforming whilst animating, and to help with the topology flow. In Maya, it's preferred that models are creating in quads, but all quads in Maya are essentially tris that have been combined. However, when doing high detail sculpting/animating, only quads should be used.


The next phase is texturing, and when that's completed, the character is handed off the to riggers/animators to add a rig to the character so that they can be animated and implemented in a game.


A point of interest of this livestream to me was when Sweeney talked about how strong concept art of a character meant his models were far greater. Character art is something I hope to create for a game one day, so it was interesting to hear how the proxy mesh can limit the design choices for a character. For example, since nothing extra can be added, things like dangling chains can't be attached to the arms. Chains could still be added to the design, by they would have to be wrapped around the arms as opposed to dangling down, which would require further rigging and animating, which will by that point have already been completed.



When creating his sculptures, Sweeney creates low poly models and then up the level of subdivisions. Above is the model he discussed in the livestream, which he used to explain how he modelled the armour from low to high poly. His method is to keep the geometry evenly spaced so that when it is subdivided, the model remains smooth and even as well. To achieve part of the model such as the helmet would require extruding the geometry outwards. In Maya, this can be done with the Extrude Tool, and the ZBrush equivalent is the Poly Extrude tool, which is selected from a menu by pressing the SpaceBar.


Sweeney's style of sculpting involves exaggerating any edges, which emphasises the edges of the armour and also adds a layer of realism to his models, as armour isn't razor sharp at the edges. In Maya, this would mean added edge loops or perhaps bevelling the edges, which would allow you to curve and add subdivisions, along with change the size of the bevel.


Another thing to consider is sculpting a face for a character, which will come from the proxy mesh. As the proxy mesh cannot be changed, you have to use something called retopology. Retopology is when a high resolution model is simplified down into a proxy model. So, using a previous version of the high poly version of the model for the face is an effective way of saving time.



If a model contains too many polygons, it won't be optimal to use in a game. If any polygons can be cut out, for example, the geometry underneath a helmet, then it should be, as this geometry won't ever be seen. So long as all the geometry needed for animations is there, there is no reason not to delete this geometry.


With models complete, the next thing to do is to UV the model so that it can be textured. Sweeney uses his awareness of the different axis of a UV map (UVW, but only U and V are used) to layout his UVs. He advised that all UVs should be straight and aligned to make sure there is no confusion with the shells. This is useful for a pipeline that has two different people modelling/creating UVs and texturing, as UVs laid out logically will be easy to understand. There are ways to straighten out UVs in the UV menu of Maya, which also includes an automatic layout tool, which takes the sizes of the the different shells into account when organising. This tool, however, doesn't take into account shells from the same object, so they can end up on opposite sides of the tile, which could be confusing to a texture artist.


UVW is similar to XYZ, which is the space in majority of 3D Software operates with. UVs only use U and V because W is the view you have of the tile, which is something I hadn't known before this livestream. XYZ also equates to the colours on Normal Maps, which usually appear as purple images with small sections of red, green and blue (RGB = XYZ = UVW). Different engines react to normals differently, and can invert them, which can usually be fixed by re-baking the maps.



As previously mentioned, high and low poly meshes can be combined in a process known as baking. Sweeney's preferred method of texturing is by using 'Spec/Glos', or Specular and Gloss maps, which are used to describe which parts of a model are shiny and to what extent. The other method of texturing is Metalness/Roughness, which uses the binary question 'Is something metal?' to determine whether something should have a shine. This method isn't used by Sweeney. Spec/Glos entails adding a diffuse (colour/albedo) to a model, and then a Gloss map. Gloss means how smooth a model is. Sweeney used an apple in his example, as apples are shiny in texture. Specular, in the case of an apple, is its reflection, which will always be grey, as it doesn't conduct electricity. A metallic object will be black, unless the metal has scratches, which will be in grey.


Another 3D Artist is Jas Dhatt, the Lead Artist at Creative Assembly. She specialises in creature sculpting, usually of reptilian characters, and has also worked on the Warhammer games like Danny Sweeney. She also specialises in design philosophy, which means taking the anatomy, characteristics and implications of a design into consideration. This means rooting the creature in realism/ the world its going to be in to make sure it fits in with the rest of the world and story. This means drawing inspiration from real-life, which can make monsters more believable and give them more of a backstory to their character/species.



As storytelling is something I'm incredibly interested in, Jas Dhatt has been a great source of information, as organic creatures is something I've never tried to model. As someone who is most comfortable creating character art, this is something I would be interested in for the theory alone.



Dhatt and Sweeney both starred in a livestream about creating creatures and monsters. They brought in the concept of anatomy, which greatly affects how a species might be shaped and operate. They used digestive systems from carnivores and herbivores to demonstrate the differences between animals, which can be used to differentiate diet, and how their anatomy is designed to help them survive. Applying this to fantasy creatures can give them a basic persona, like a creature with eyes on the side of its head to see predators coming. Bone structure needs to be examined as well, as some creatures have evolved with bones to support different muscles depending on what the species does.



The main things to take into account when creating a monster are flow, balance, shape language and forms. Flow is a lot like leading lines in a painting, and are used to lead to the focal points of the creature. Its also about making a creature look organic, which means avoiding right angles and parallel lines, which are too geometric to be believable on a creature.



Balance is important in any design. This can mean a lot of things, the most simple being having an easy to read design, and colour. Colours are usually used in a rule of three, with one colour being the main one, then another colour used slightly less on the design, and then one or two colours for the accents of the design. Balance can also be used to give a drawing more character. Sweeney posed the idea of using percentages to work out how a design should be. For a rough, spiky character, 70% of the lines should be rough, and 30% should be smooth. This can also create focal point on a design.


Imbalance in design can be unnerving, which can be used in more horrifying characters for dramatic affect.


Shape language is about using shapes to create emotion. Pointy shapes invoke a sense of danger, whereas rounder, spherical shapes are softer and more timid. In short, long, straight lines are used on more intimidating monsters, and softer, curved lines are usually more prominent on harmless-looking creatures.


Finally, form is the creature's silhouette. The silhouette must be recognisable, and can also be used to balance out the placement of colour. With all of this in mind, designing a creature is made a lot simpler, as there are separate things to research instead of just trying to jump straight to the designing phase of creation.


Sweeney uses Autodesk 3Ds Max for his modelling and sculpting, as well as Marmoset for his texturing, which are two programmes I'm unfamiliar with, but share similarities with other software. For example, Maya, Mudbox and 3Ds Max are all owned by Autodesk, which means some of the interfacing and tools will be similar, meaning that there's a good chance I could learn how to use 3Ds Max a little faster than someone who'd never used an Autodesk programme before. This also means that the processes Sweeney used can be easily translated over to Mudbox and Maya.


Overall, these two livestreams have taught me a lot about both the technical side of modelling and the design process for a character, as some of the concepts of designing a creature could be applied to other things, such as environments. Seeing how these two brilliant 3D Artists work and think has also broadened my horizons, as the processes they use to achieve their models are things I hadn't considered before, like making proxy models and understand muscle structure in different animals due to their behaviour and how that would affect their appearances. Both of these livestreams have taught me a lot more than I knew before, and this knowledge will be used in the future.



Sources:


Comments


bottom of page