With the sculpt completed I moved on to retopology, creating a simplified mesh over my high-poly sculpt that is greatly more optimised and would be much lighter to import into a game and run continually on a machine. To ready my model for retopology, I first decimated it to lower the overall polycount and redistribute the faces. This was able to be done in ZBrush. I then exported the model as an FBX file for import into Maya. During this process, I removed the fabric clothing on the model, as I felt this would be too difficult to retopologise and would overly complicate the process. As such, when importing my sculpt into Maya, it came with in three Objects: the hair, the body and the eyes.
When importing my model into maya it came in with an incredibly bizarre texture, almost looking as if two or more models of different materials were intersecting or clipping together at once. I had no idea what was causing this, though Kenny said it was likely because - at some point in the sculpting process - I merged multiple subtools together atop one another. This confused me, as the model looked and manipulated fine in ZBrush, and ZBrush never gave me any indication that such an error was made. Additionally, if this was an irreparable error, it is likely that I would have had to completely re-do my sculpt, starting over from importing the base mesh into ZBrush again. This strange quality made the initial retopology process incredibly difficult and time-consuming, as - for some reason - the Quad Draw tool would not work in certain areas, even though the entire mesh had been set to Live. Additionally, the Quad Draw remesh was being drawn beneath the model, making it nearly impossible to see where retopologised faces sat in the mesh.
Whilst I was struggling through this process, I had an idea: that if this had been caused through previously merged meshes intersecting, perhaps I would be able to separate them again. I selected the base sculpted mesh and clicked "Separate", and indeed the intersecting meshes did separate, allowing me to delete the erroneous one and continue working on the sculpt. This greatly increased the ease in which I was able to retopologise, and likely shaved hours off the total time it took. This also let me properly use the Quad Draw tool and draw quads anywhere on the form of the live base mesh.
When retopologising, I was advised to separate disparate parts of the mesh: to work on them as different components as opposed to one whole. As such, I decided that I would work in different sections of the model: on the torso, the legs, arms, hands, and head all separate. This could be done through the Layers function of Maya. When retopologising, then, I worked to create concentric rings around the model, with quads looping and wrapping around the form in a consistent flow. This would hopefully make the rigging and weight painting process more manageable, though it is likely that the retopologised mesh is highly unoptomised, as I found it difficult to gauge where areas of stillness and detail should sit. There are a lot of curves in my character, especially in the bulky torso and back, and as such I tried to use edge loops to create areas of higher geometry where the quads clutched tighter to the character's form. However, I was worried that, when it came to weight painting, some of these areas would not deform well, as often these quds were of drastically different sizes.
With the torso done, I moved on to the legs. This was a similar process, though during this, I worked in longer shapes that ran down the length of the leg and thigh. Again, I found some of the placement of quads difficult, as I wished to keep the detail in the knee and calf regions of the model. During this process I also retopologised the foot, which I kept as a part of the leg and not its own separate mesh due to the lack of detail and flow the two masses have between them.
As stated previously, I kept the different regions of the model in different retopologised meshes. I accomplished this through the use of Maya's layer functionality, with different layers named appropriately to the body part they represented. I also appropriately named the different meshes in my outliner to keep them organised and clear.
The arms, hands, head, hair and eyes then followed a similar process: I worked in concentric rings that wrapped around the forms of the model, using edge loops to add geometry. In the event that these retopolgies do not deform well when rigged and weight painted, I will likely come back to them to add more geometry and reorient faces. I am definitely not confident in my ability to retopologise, as this is more or less the first model (and certainly the first of this scale) that have worked on. Nevertheless, I did not find that the process disagreed with me particularly.
With the retopology complete, I sought to UV unwrap the mesh. This would allow me to texture the low topology mesh, as well as hopefully bake the detail of the high-topology mesh on to the lower detail mesh through the use of a normal map. Again, I am unsure how well such a process would go, as I wonder whether or not there is enough geometry in the retopologised mesh to support normal map baking.When UV unwrapping, I sought to mark "seams" in the geometry of the model that would allow the 3D mesh to be unravelled into a 2D plane. Like with the retopology process, I worked systematically as I UV unwrapped, working on every mesh as a separate piece and placing all unwraps into the same grid square.
During this process, I tried to keep a consistent flow across the entire surface of a model and minimise warping between different areas of the checkerboard pattern. Some warping does remain, however, in areas of sharp pinching, such as around the breast on the torso.
With the UV unwrap complete, I will next move on to baking, texturing and rigging the model. I am pleased with the progress I have made and feel that I am on track to get, at the very least, the model fully textured by the deadline. In session time, from here on out, I will likely work to complete the rigging and weight painting of the model, as these are elements of the process I feel the least confident and experienced with.
Comments
Post a Comment