3D Software Techniques

For this project, I will need to create some kind of environment/asset in a 3D software. We will be using blender, and so I have specifically researched into blender, and the different tips and tricks on how to model and texture effectively, as well as the different ways that one can model different kinds of assets.

Modelling Techniques

There are so many different tips and tricks to be found online regarding modelling, specifically in blender. I will go into more detail about all of the different techniques individually, however here is some general information that I thought was useful to know about:

There are various useful key bindings and shortcuts that blender and other 3D modelling software use for editing; the most common ones to know are G to move, S to scale, R to rotate, and E to extrude. These shortcuts can be combined with numbers and axes to specify which direction and how much one would like to edit.
By smoothing the surface through the use of shading or subdivisions, you can get some insight into how your asset will look once fully rendered, and so by utilising this you can model appropriately, and smooth out any issues early on rather than waiting for it all to be done and then realising your mistakes.
There are a lot of different general tips when modelling; for example, it's a good idea to have "merge by distance" turned on when modelling, to clean up any small holes or cracks in your mesh. It's also generally known that when modelling, especially when box modelling, you want to make it out of quads, or shapes with 4 sides. This is so that it is easy to perform some of the more advanced actions, such as loop cuts or ring selections. Triangles will also mess up any subdivisions you want - this is one reason why it's always better to use a UV sphere rather than an ico sphere.

Box/Polygon Modelling

Box modelling is probably the most common type of modelling technique in blender. This usually starts with a "primitive mesh", (e.g. a cube, sphere, or torus), and then you are able to model a complex shape. These meshes can be modelled in edit mode, object mode, or even sculpt mode. A more advanced form of box modelling would be poly modelling, which is where a mesh is created one polygon at a time - this gives the creator more control over how large each polygon is, and the placement/proportions of the mesh before continuing to model.

Polygon modelling is very similar to box modelling, except instead of starting with a 3D primitive mesh, you start with a single vertex or edge. This method puts more emphasis on building your mesh from edges and vertices, and getting more detail. the advantages and disadvantages of this method are pretty much the same as box modelling.

This technique has many advantages - it is very versatile, and is a sort of stepping stone to the more advanced modelling forms. Anybody looking to learn modelling in blender should start with a box. One disadvantage of this technique is that it's more difficult to get organic-looking shapes by starting with a simple mesh, without things being distorted by subdivisions and smooth filters.

Curve/Surface Modelling

These are forms of modelling that I will probably not be using - it is all about algorithms, mathematics, and circles. Curves refer to flat edges, which can be edited similar to vector lines in Adobe Illustrator, for example. There are several different kinds of curves, such as Bezier curves or NURBS, however what sets curve modelling apart from other kinds of modelling is that you get complete vertex control.

A "surface" is basically just an extended curve, and is a 2D plane with only two dimensions - U and V. With this type of modelling, the creator gets intricate control of these dimensions, and can create curved, dynamic shapes. The upside of this kind of modelling is that it allows you to create completely curved shapes, whereas with box modelling your shape is entirely comprised of polygons - this can be useful for controlled hair/fur. The downside of curve/surface modelling is that it can get quite chaotic quite quickly, and often involves a bit more mathematics than any other kind of modelling.

Sculpt Mode

This is a common type of modelling used in blender to create organic shapes, such as humans or animals. The difference between this kind of modelling and others is that instead of working with sharp, shape-based primitives and manually pushing and pulling the faces around to create an asset, you use brushes to mould the asset organically.

A good thing to note about sculpting in blender is that there is this technology called "multi resolution", which allows the modeller to create new levels of detail and store the mesh between each layer, allowing the model to get more detailed as the modeller needs, instead of having to work with thousands of polys right from the beginning. This also prevents the model from lagging out the software and crashing due to too much complexity. This is something that I am definitely going to be looking further into as time goes on, since I am a huge fan of organic modelling and the human form.

Photogrammetry

This is a unique type of modelling, in which an artist takes many pictures of a real life object from many different angles, and uses it as a reference for their desired mesh. The software attempts to replicate not only the shape of the real-life asset, but also the lighting and texture of it. This can be used to create realistic textures, however requires a complex UV map.

This method has lots of upsides and downsides, most of which are fairly obvious. The pros are that it can create complex meshes and textures very quickly, and speeds up the initial modelling process. The downsides are that it requires a lot of clean-up; this method captures not only the desired asset, but also anything surrounding it (e.g. ground, debris, etc.) which can be annoying and complex for more organic assets. For example, if you wanted to photogram a tree in a forest, you would have to edit out the surrounding floor and leaf debris. Other cons include the fact that you have to have the object in the real world in order to do this method; fantasy objects/ faces would be harder to do, and would require more adjustments. And, of course, any surface on the object touching the floor/surface while the photographing is taking place would not be rendered.

Simulation

This is an extremely unique and technical type of modelling, more often used in the context of 3D animation. There are many different kinds of simulations in blender, for example physics, cloth, soft body, water, fire/smoke, and particles. These can be used to create physics-based assets (such as a cloud, or water splashing against a rock) without having to model it by hand. This method uses parameters that applies to real world mechanisms, and so can create some interesting effects when fiddled with.

The pros of this method are that you can create dynamic, realistic scenes with minimal effort. The cons is that it's a fairly complex method, and can easily get out of hand. It's also quite difficult to control the exact shape of these models without losing the organic properties of the asset.

Procedural Modelling

Procedural modelling is a type of modelling which can help create lots of different variations of different assets. you can make your own procedural brush, or take one from online, however the defining feature of this particular kind of modelling is the ability to create similar but distinctly unique assets from simple mathematical and aesthetic parameters. Examples of parameters could be number of floors in a building, the direction of an extrusion, or the shape of a leaf.

Procedural modelling can be applied to both geometry and meshes, as well as shading and texturing. Procedural shading is linked closely to displacement, and by using a displacement map one can take a primitive mesh and turn it into something random and unique. The cons of this kind of modelling is that it is very mathematical, and creating a procedural brush is fairly complex, however the payoff is a greatly reduced workflow if you are looking to make a forest, or a large amount of debris, for example.

Boolean Modelling

This is a type of modelling that is very closely linked with box modelling, and is often used alongside it. Using this method, one can cut out whole chunks of a mesh very easily. There are three different kinds of Boolean modelling; difference, union, and intersect. With difference, you take out the shape and volume of one mesh using another one, so that the cut-out becomes negative space rather than positive space. With union, you merge two meshes together, and with intersect you keep only the geometry where the two overlap.

This is a fairly simple type of modelling, which is mostly used alongside previously mentioned modelling techniques, and so I will probably end up using it in some form or another at some point during my production process.

References

Texturing Techniques

Texturing in Blender, it seems, is fairly simple. With free online resources, one can create an effective shaded texture quite easily - you just have to have a decent understanding of texture roughness, normals, and displacement. I decided to try out some texturing in blender, using online tutorials, to try and figure it out. I also wanted to specifically try some of the more complicated methods of doing texturing before I begin trying out the "budget" methods.

First of all, I went online and found 3 basic textures to use as experimentation. I began with texturing a simple cube with a wood texture - I wanted to fully understand how each of the nodes in the texturing interface work, and what PNG translated to what aspect of that texture. For example, I learned from my basic sources that the "colour" PNG of any texture connects directly to the "base colour" input of the texture. Along with this, the "roughness" connects to the "roughness" input, and the "normal" connects first to a "bump" node, and then to the "normal" input. Finally, the "displacement" image plugs into the "displacement" input of the final output node. This texture did not include it, but if there was a "metallic" PNG, it would go into the corresponding "metallic" input. Colour controls the overall colour and pattern of the texture, roughness controls things like the shine of the texture, and the normals and displacement dictate how much the texture appears to be extruding. This is the simpler way of making an object appearing 3D, and interactive with the environment.

The second object I textured was a torus (donut-shaped mesh) with a marble texture. For this, I decided to start fiddling around with advanced displacement maps. I read up on them online, and found that if one wanted to create a realistic displacement map, which actually distorts the mesh itself, you need to first create a ton of surface subdivisions in order to smooth out the shape and create lots of vertices, and then apply the displacement PNG to a "displace" modifier. To see a simple step-by-step guide on how to do this, read this guide I found online. This is a more advanced way to make textures look realistic, although I have been informed that it is not very practical for larger/more complicated assets, as not only does it distort any desired shape/size that you model this object to be, but the increased number of polys all controlled by an image can crash blender quite easily (even with the small amount I was experimenting with these effects, I noticed a significant decrease in fps as it got more complex).

This became ever more apparent when I tried to texture a sphere (which I set the subdivision level viewport to 6) with a complicated rock texture. This very nearly crashed blender, and so I think that while using an advanced displacement map is fun and dynamic looking, for my final assets I might just stick to normals. (The image for this rock texture unfortunately corrupted, so I cannot display it).

Referances

Page updated

Google Sites

Report abuse