Sometimes it’s easier to create an object’s basic form using one method—for example, sculpting or combining simple primitives—before creating the final topology for texturing and the final renders another way, using the basic forms as a guide. This is what retopology allows us to do. Retopology is the process of creating new geometry over an existing mesh while maintaining the object’s basic shape and rebuilding its topology. Retopology allows us to split the tasks of creating the right shape and creating helpful topology, making it easier to produce the model we want.
In this chapter, you’ll learn how to use Blender’s Snapping tools and other retopology techniques to model new geometry over the surface of highpoly, sculpted meshes, like the ones we sculpted in Chapter 6. The result will be a final mesh that captures the details of the sculpts with fewer polygons and produces topology that subdivides well and is suitable for rigging and animation.
There are several ways to retopologize, but the most direct is to use Blender’s projection Snapping tools to create geometry directly on the surface of the mesh you want to retopologize. (See Snapping for more on these tools.)
To start retopologizing a mesh, take the following steps:
Add a new plane object in Object mode.
Enter Edit mode and turn on Snapping.
Change the snapping target to Faces and enable Project onto Surface (see Figure 4-10) to make Blender snap selected vertices and edges onto the surface of other objects by projecting them from the current viewpoint.
Scale down the plane until it is the same size you want the faces to be in your final retopologized mesh. Next, grab and place the plane over a region of your mesh where you wish to start retopologizing.
Adjust the positions of the existing vertices. Then, select one edge of the face and start extruding out to create new faces, following the contours of the mesh.
Whenever you extrude a new face or grab an existing one, it will be projected onto the surface of any other mesh on screen (though other objects you selected before you entered Edit mode will be ignored).
Using the method described above, in a short time you should be able to cover the surface of even a complex mesh with the exact topology you want, creating a new mesh with the same shape as your original.
There are other useful ways to retopologize quickly that are worth discussing. The first is to use the Shrinkwrap modifier to project simple topology over the surface of a mesh. With this method, you first model a basic cage around the outside of your original model (as shown in Figure 7-1), extruding areas to create the topology you want. The cage needs to match the original shape only roughly because the next step is to add a Subsurf (or Multires) modifier and a Shrinkwrap modifier to the mesh, setting your original model as the target for the Shrinkwrap modifier. The Shrinkwrap modifier then automatically conforms the new mesh to the surface of the old one.
Shrinkwrap has a few methods for doing this, and it’s best to experiment to see what works best for your model. I’ve generally found the most useful methods to be the Nearest Surface Point and Project modes. Note that there are some extra options for Project mode and that you’ll need to turn on both Negative and Positive in the direction checkboxes in order for your mesh to project in both directions to conform to the target (unless the surface of the new mesh is always above or below the surface of the target).
This method of retopologizing works well when you want to quickly retopologize a sculpt that needs a better base mesh—for example, if you started a head or even a whole body from a cube and then wanted a base mesh with topology that does a better job of supporting the basic forms.
The Shrinkwrap method is also useful for combining models composed of multiple pieces into a single mesh. For example, to create a one-piece version of the Mayan glyphs discussed in Chapter 6, you could project a simple grid mesh onto the glyphs, as shown in Figure 7-2. Notice that I’ve repeated the Shrinkwrap and Subsurf modifiers to capture all the details. This repeating of modifiers is a useful trick, especially when you are projecting onto a mesh with a lot of sharp corners or deep creases and folds, because the first shrinkwrap will miss details that aren’t visible when looking straight at the surface of the target mesh. The resulting mesh can then be easily sculpted or unwrapped and textured without having to work with multiple part meshes.
Another method for retopology uses Blender’s Bsurfaces add-on, which has some neat tools for retopology that allow you to draw lines with the grease pencil and have them automatically converted to a mesh, as shown in Figure 7-3.
I created most of the Jungle Temple project using traditional modeling methods, without a lot of retopology. However, when blocking out the trees, I came up with nice placeholders by snapping curves over the surface of the background objects. These placeholder curves can easily be turned into more finished meshes using retopology. To apply this technique, take the following steps:
Create a duplicate (unlinked, SHIFT-D) of the tree placeholder. Then, convert it to a mesh with ALT-C▸Mesh From Curve/Meta/Text. This creates a mesh with roughly the topology we want, except that the roots and trunk are separate meshes.
Join the roots to the trunk by deleting the vertices in the areas where the two should join. Then, start creating new geometry to cover the join and bridge the gap (see Figure 7-4). Select pairs of edges that roughly line up on either side of the join and create faces between them (F).
You can then add some loop cuts (CTRL-R) across this face to split it into more even-sized quads.
Grab the new vertices this creates and release them again to snap them to the surface. Repeat the process to continue filling in the gaps, joining new faces together and trying to keep the topology as even and grid-like as you can.
Repeat the above steps for all of the joins between the roots and the trunk.
Finally, close the open ends of the roots. It’s also worthwhile at this stage to adjust the density of edge loops around the roots and trunk, deleting some (X▸Edge Loop) in areas where they are too dense and adding them in sparser areas (CTRL-R).
While the original Bat Creature’s base mesh was a good start, it lacked the necessary topology to support all the features of the head and more detailed anatomy. Additionally, the wings were initially constructed and sculpted as separate pieces. Now, I needed to create a final mesh with more detailed topology that combines the two parts.
Before using retopology, I needed to deal with the fact that the full sculpt was in the millions of polygons and was getting unwieldy. Of course, this was the reason for retopologizing it, but still, it would be good to have a stand-in mesh that captures the details with a lower polygon count for me to draw my new topology over. The answer is decimation.
Decimation is a process whereby a high-poly mesh is automatically simplified by collapsing small edges and polygons to reduce the model’s complexity while losing as little detail as possible. Blender has a Decimate modifier, but it’s a little slow and not always effective, so I often use another open source application called MeshLab (http://www.meshlab.sourceforge.net/). To use MeshLab on a Blender sculpt, take the following steps:
Export your object as a wavefront object (.obj) file (File▸Export), making sure that the Selected Only and Apply Modifiers options are checked so that all modifiers are applied (including Multires) and that the details of your sculpt are preserved. This will create a fairly huge .obj file that contains your high-poly sculpted mesh.
To import your sculpt into MeshLab, just run the program and then use File▸Import to import your mesh in .obj format.
MeshLab has a huge array of options for processing meshes, but for our purposes, we’ll use the Quadratic Edge Collapse Decimation tool (Filters▸Remeshing▸Quadratic Edge Collapse Decimation). This brings up a menu allowing you to specify a target polycount for your mesh, among other options (see Figure 7-5).
Set the target poly count to around 150,000 and turn on Planar Simplification in the options. Then run the filter. After a short wait, you should be presented with a much lower polygon count version of your sculpt that still encompasses much of the detail. You can export this from MeshLab using the File menu and then re-import it back into Blender.
With the sculpt decimated and re-imported back into Blender, I could start laying some new topology over it. I began by adding a new plane in Object mode and applying a Mirror modifier to it. Then, I turned on Projection in the Snapping tools buttons. Next, I grabbed my plane in Edit mode and scaled and positioned it where I wanted to start my retopology from (the torso, in this case).
At this point, I thought carefully about what density of polygons I wanted in the final model. Scaling the first plane up or down and working to this scale made a big difference in the final poly count of the retopologized mesh. I generally tried to work at a similar scale to most of the important details of the original mesh.
With a starting point set, I began extruding (E) from one edge of my original quad along the forms of the model (see Figure 7-6), following the outline of the ribcage and torso muscles. This created a face loop that ran along an important form within the model. By filling in the surrounding areas and creating further face loops along other forms, we can work toward completely retopologizing our sculpt with topology that supports the forms we have sculpted. This creates a more efficient mesh for capturing the fine details when we subdivide it and makes it easy to unwrap and rig if we desire to later.
As I progressed over the model, many of the same general positions for face loops that I used in the base mesh were useful again. While overall the topology was more complex, creating face loops that ran around the armpit and shoulder and around the arms, legs, and neck resulted in clean, easy-to-modify topology (see Figure 7-7). Also, because the mesh was still broadly symmetrical, I only needed to retopologize half the body and use a Mirror modifier to fill in the other half.
For some areas, like the face, I began to shrink the polygons in order to pack more into the area’s smaller, more detailed forms. I created loops that ran around the eyes, nose, mouth, and outside of the ears (see Figure 7-8). These loops both supported the forms of the face so that they weren’t lost or softened too much when subdivided and added extra density to these areas to allow for the addition of more details with the Multires modifier. (For specific tips on face topology, see Head Topology.)
To create a mouth cavity, I extruded back from the edge loop surrounding the mouth and filled in the hole (see Figure 7-9). When we return to sculpt on this retopologized mesh, we’ll need to resculpt this area, but the mouth will now be able to be opened and posed.
For objects like the wings, where it can be difficult to see into tight areas, it can be handy to switch to perspective view (5), which allows the camera to be moved in and around tight spaces. It can also be useful to restrict the view with ALT-B. This lets you select a part of your model to show in the 3D Viewport, allowing you to see a small area without other parts of your model getting in the way, as shown in Figure 7-10.
Creating new topology that bridges the join between the wings and body is as simple as continuing to retopologize over the join (see Figure 7-10, which shows the mesh covering both the wing and shoulder of the Bat Creature). Doing so allows you to create a single mesh that combines the wings and body (see Figure 7-11).
My goal for the Spider Bot was to turn the rough sculpt into a smooth, hard-surface model. To do this, I needed to replace the original, simple topology of the base mesh with a denser mesh, tailored to the shapes of the body. Additionally, retopology allowed me to embellish as I went along, turning parts of the model into hollow shells, adding holes, or adding supplementary details to the retopologized surface.
Beginning with the body, I added a plane (SHIFT-A) as before, with Snapping turned on and set to project new geometry onto the surface of the sculpt. Next, I added a Mirror modifier to my plane and began following the major lines across the surface (see Figure 7-12). The final step was to fill in quads to complete the rest of the surface, while trying to keep the size and distribution of polygons even. To smooth out any dents in the surface, you should select an area you’ve already retopologized—avoiding edges and corners—and use the Smooth tool a few times to even out the distribution of geometry.
For some parts of the sculpt, I wanted to replace the rough forms sculpted freehand with more precise geometry. To do this, I simply added a primitive, such as a cube or cylinder with approximately the right shape, and moved it into place. For example, in Figure 7-13, I added cubes to form the sockets for the legs, deleting their front and bottom faces and subdividing them to give the right shape while adjusting the surrounding geometry to fit them in.
To further refine the model, I added support loops around its main hard edges (see Figure 7-14). You can do this using tools like Loop Cut (CTRL-R), Subdivide (W▸Subdivide), and Edge Slide (CTRL-E▸Edge Slide), or you can use the Inset Faces operator, which adds an edge loop around the selected faces with a constant thickness. Combining these operators lets you build up support loops around the key hard edges of your model.
You can add embellishments to the retopologized mesh in several ways. For example, the piece on the left in Figure 7-15 was created by duplicating (SHIFT-D) some of the faces of the retopologized mesh in Edit mode and moving them up (with snapping turned off) to create a new piece. Then, by extruding the new part to give it some thickness and adding some support loops around the edges, I created a raised area on the surface of the model. The circular element on the head in Figure 7-15 was made by creating a new circle, projecting the circle onto the surface of the model with Blender’s Snapping tools, and then turning off snapping and extruding out to create a raised area. To flatten the top of it, I scaled the central part along its normals by changing the transform orientation for the 3D manipulator widget to Normals and then using S▸Z▸Z▸0 to scale the selection along the surface normals to zero, making it perfectly flat.
I repeated the same process for the abdomen (see Figure 7-16) and the legs (see Figure 7-17). In some areas, my sculpt had lumps or areas that didn’t quite polish well enough while sculpting, so I had to tweak things a bit once retopology was finished in order to get nice clean edges.
Because I sculpted only a couple of variations for the different leg pieces, I had an easy time when retopologizing. Each unique piece was retopologized only once (as shown in Figure 7-17) and then duplicated as linked duplicates to create the copies needed for the legs. As in the case of the body, I also used a Mirror modifier to mirror the leg pieces across their axis of symmetry.
Because I didn’t really need to copy the legs to the other side of the body properly at this stage, I added all of the leg pieces to the left side of the body as a group (CTRL-G in Object mode) and then created another instance of this group (SHIFT-A▸Add▸Group Instance▸Name of group). The result was a noneditable, duplicate instance of all the objects in the group. I then scaled this group by –1 on its x-axis to flip it to the other side of the body. Later, I could use the Make Duplicates Real operator (CTRL-SHIFT-A) to convert this group to real geometry (still as linked duplicates of the original objects), but for now, this was a convenient way to preview the entire model without creating unnecessary objects (see Figure 7-18).
In summary, here are some general tips to follow when retopologizing your models:
Model the important edge loops first and then fill in the rest.
Use support loops or creasing to tighten edges.
Use Blender’s retopology tools to add extra embellishments to your model.
Try to keep an even polygon density, unless you are specifically adding more detail in a certain area.
Pay attention to the silhouette of your model. Usually you will be looking at the area you are retopologizing face on, so be sure to look at it from other angles to see that it works.
Try to work with larger polys where possible. Don’t use more than you have to!
The head is such a complex subject that it deserves some special attention here. While there is no one topology for every conceivable head, there are some important principles to consider. Mainly, when retopologizing a head, you should concentrate on creating face loops around the key features of the face, particularly the mouth and eyes, which will make it easy to deform the face into familiar shapes (see Figure 7-19). These principles apply whether you are retopologizing a sculpted head or modeling one from scratch.
By constructing the important areas of the face with nice topology first, it’s easier to join the rest of the head together. In general, I begin with the eyes and work down from the nose into the mouth. Then, I work backward through the cheeks, forehead, and ears. Lastly, I cover the rest of the head and on down the neck. By first building outward from the most important and complex parts of the head, you’ll have to sort out fewer problems with the topology in the areas that matter most.
When creating or retopologizing a head, there should almost always be a clean ring of face loops around the eyes (see Figure 7-20). These face loops should extend out from the outline of the eyelid to encompass the brows, the outer area of the bridge of the nose, and the upper surface of the cheekbones. This ring will make it easy to close the eyes or raise the eyebrows or cheeks, and it mirrors the underlying anatomy of the face: The orbicularis oculi muscle surrounding the eye circles the eye in exactly the same fashion as this ring of face loops.
As with the eyes, the orbicularis oris muscle encircles the mouth and is responsible for widening and narrowing it (see Figure 7-21). To make it easy to produce common motions of the mouth, we’ll create a ring of edge loops around the mouth in the same way. This ring will also make it simple to define the outline of the lips and deform them easily.
The complex features of the nose are the nostrils and its tip. By incorporating face loops that run around these areas, the rest of the nose becomes relatively easy to define.
The nasolabial fold is an important feature of the nose and is most obvious in snarling or older faces. Adding a face loop that runs over the bridge of the nose and down the sides of the mouth, connecting under or at the chin, allows us to define this area (see Figure 7-22). The bridge of the nose is formed simply by bridging across between the loops encircling the two eyes.
Ears are highly variable from person to person, but their overall construction is reasonably constant. Because the ear is mainly cartilage with no musculature or articulation, it is a pretty static feature of the head and thus a good place to hide awkward triangles if necessary. By defining the helix and antihelix (the outer and inner curved parts of the ear, shown in blue and green respectively in Figure 7-23) with edge loops that run down into the earlobe, we can make it easy to define the ear’s overall structure. The other main feature of the ear is the ear canal. In addition, it is sometimes handy when constructing the ear to try to have a single edge loop encircling the ear to make it easier to attach the ear to the head. One way to do this is to construct the ear as a separate mesh initially and then create a loop around the outside before placing the ear and connecting it with the head.
In this chapter, you’ve learned how to retopologize meshes with arbitrary topology, be they high-poly sculpts or a collection of separate primitives, to create smooth, clean topology that subdivides well. We’ve discussed ways to use Blender’s Snapping tools to retopologize meshes, as well as alternative methods, such as the Bsurfaces add-on and the Shrinkwrap modifier. We moved on to using these tools to retopologize the meshes we created in earlier chapters with new, better topology.
In Chapter 8, we’ll unwrap this new topology to provide it with texture coordinates before baking and painting textures for our final models in Chapter 10 and Chapter 11. The improvements that we’ve made to our models in this chapter will make that process easier and will also improve our render times when we reach Chapter 14.