9/3/20
This series of blogs will serve as a record of my progress for my VIST 372 projects. I will aim to describe my progress throughout each assignment with as much detail as possible.
The Live Action Set Extension is the first project of this course. The goal is to recreate a still (approved by the professor) from a black and white movie filmed before the 1950s by replicating certain set pieces in CG, and then seamlessly compositing the CG elements onto a real-life photograph, AKA a plate. One restriction with this assignment is that we are to use NURBS (Non-uniform rational b-splines) to model.
For this project, I am using Maya to recreate a still from Charlie Chaplin's The Kid (1921).
From this image, I am choosing to model everything except for the alley behind the arches.
With this particular shot, I thought it would be a challenge to model some of the more detailed components, such as the window and the door, but then I discovered that the real challenge was learning how to model with NURBS and applying those techniques effectively and efficiently. Coming into the project, I knew very little about NURBS, and I quickly learned I had hardly scraped the surface of what they are capable of (before, I had only ever used them for character rig handles). As you can imagine, I was quite intimidated from the start, so I began modeling with what I knew, using polygon primitives. My plan was to start getting the basic shapes down with polygons before switching them to NURBS later on.
I modeled out the basic shapes using normal polygons, using my still as a reference. I used multiple polygon cubes, inserted edge loops to mark where specific components were going to be and after a series of extrudes, I got the basic shape of the building and arch.
My next step was to create the poles. These were the first objects I modeled using NURBS. I traced the outline of the pole with the EP Curve tool, revolved, and duplicated.
After the poles, I focused my attention on the window. Starting from another polycube, I matched it up to the window indent in the wall, and used a series of edge loops and extrudes to make it look right.
For the door, I made a cube and inserted edge loops to mark out where the specific details were going to be placed before I model them out later on.
In the meantime, I decided to try out the method of converting polygons to NURBS.
I started by making a duplicate of the wall to test it out. I converted polygons to subdivisions, then subdivisions to NURBS. This resulted in my object being shaded bright green and covered with a big mess of fragmented geometry. The model was split up into a total of 51 different sections. If one section was selected and moved, it would move independent from the rest of the model.
The good news is that the issue with the green was resolved quickly after I assigned the default material. The bad news is I am not sure what to do next. I do know that the smoothness of the model is due to the lack of bevels in the edges, but I am not sure how to bevel with NURBS. I tried beveling the edges in the original polygon model of the wall before converting back to NURBS, but that result was only slightly better, but also worse than before.
At this current point in time, I am not where I should be at in the project, but I am determined to fix my problems before I move on to the texturing phase. I will start by learning the correct process to convert polygons to NURBS without breaking the model, and I will start modeling the details on the door.
9/8/20
I set aside the NURBS conversions issues for a long time to figure out how I should approach doing the designs for the doors. I found the full movie on Youtube and found the same door in closer detail for reference.
My first approach was to use boolean commands to carve out the designs on the door, but I found this to be difficult because it was nearly impossible to be precise. Certain design elements weren't lining up properly, not to mention it was a tedious effort.
My second approach was much more efficient. I used two NURBS circles and manipulated them together to create a sort of cross-section of the door design. This would be duplicated 3 times to create four of the same cross-sections. I placed them in a rectangle formation, rotating them accordingly, and then I used the loft tool to create surfaces from corner to adjacent corner, as shown at the right.
To create the innermost door design, I used the EP curve tool to draw out a more angular cross-section. Then I repeated the process above to finish out the design. Initially this left a hole in the middle, so I just lofted the isoparms of the left and right inner elements to fill it up.
I repeated these procedures to create the longer design elements in the upper area.
For the numbers and the horse shoe decoration, I drew out the shapes with the EP curve tool, then I placed a NURBS circle at the ends of each one, and used the extrude tool. Then I manipulated the control vertices to close the ends up.
For the the sheet on the door I used a manipulated a NURBS plane and the doorknob, I used a simple NURBS sphere. Combining all the elements together and I got my door.
After this, I turned my attention back to the matter of converting the current polygon methods. After communicating back and forth with Jorge the TA, I realized I was complicating things by trying to convert the polygons. Instead, he suggested using booleans to cut out the necessary shapes, such as the openings for the windows, the door, and the arch. However, this didn't mean converting to NURBS wasn't going to help, nor did it mean my original polygon models were completely useless. The original polygon models were now going to be used for reference.
I started with some of the easier elements. I just made multiple NURBS cubes and matched them up to the steps and the porch in front of the door.
I made another NURBS cube and matched the size of the reference wall. With the window and door, I wanted to be as precise as possible with the placement, so my plan was to start with polygon cubes because it would be easier to alter and match up to the original polygon wall. Then I would bevel the edges closely to retain its shape when I converted the cubes to NURBS. Then I used the booleans difference tool to carve out those holes.
For the arch, I made another polygon Cube and match it to the shape and size of the arch in the reference model. I beveled the edges accordingly and then converted to NURBS to use the boolean operation
At the moment, everything is a NURBS object except for the window, the window's outer frames, and the doorframe. These should be modeled quickly and then I can finally move to the texturing phase.
After some more work, I finished modeling the base elements in NURBS.
Using NURBS cubes, I used booleans to cut out the holes for the window pane and then duplicated it.
I used another set of NURBS cubes lined up to the window and door holes and used booleans to create the window and doorframes.
For the curtain behind the window, I used the EP curve tool to draw a curve that represented the subtle folds of the curtain. I duplicated these multiple times and made minor adjustements to vary them. Then I used the loft tool to build the curtain.
Once I had all these components, I put them all in their appropriate places. I also organized the outliner to be more coherent and easy to understand.
Now to start surfacing in Substance Painter.
9/10/20
In this final update of my Live Set Extension project, I will be going over the steps I took to complete this project, as well as the many mistakes I made in hindsight.
Since the previous update, I realized that Substance Painter was not the best way to go to make textures for NURBS objects. NURBS objects do not have UVs so when I tried to import one into Substance Painter, I was surprised to learn it would not appear.
So I started asking around and looking at tutorials and I eventually decided the best option, considering the little time I had left without losing too much sleep, was to find textures online and edit them to match as close as I could to the reference image, and then assign them to its respective object.
When I had all the textures I needed, my process was this:
Right-click on an object and assign a new material. Select aiSurfaceShader, and then select the checkered icon next to image. Then I would select the file tap and select the appropriate texture I found for that object. Then I would set the roughness accordingly, since by default it is set to be smooth. None of my objects were extremely reflective so they all had to be changed.
It took a while to do this for every object, because for some reason, in this one day of my working on this project, Maya had a fatal error more times than ever in my history of using it. Once that was complete, my next step was to line up the model to the plate I was going to use.
I made a new camera and assigned the above photograph as the image plane. I set the resolution gate and render settings according to the resolution of the original reference. Then I actually started lining image up with the model. It was in this moment I learned a few things. The first problem I noticed was that my arch was a wee bit too wide, and I thought that was it. But with more fidgeting, I found that my plate and my model would not line up the way it does in the original reference picture at all.
This was because the plate was shot terribly. Although it was overcast when I went out to take the pictures (dozens of them), and the location itself seemed plausible, I failed to take pictures at the proper angle and level. The shot I was recreating had a very obvious right angle, and so did the plate. The unfortunate thing was that these angles did not line up properly because I shot the picture too high.
After some deliberation, I decided to continue as normal, because there wasn't any better picture to use as a plate.
I set up a dome light to create the ambient lighting, then I used a spotlight and edited the roundness setting to create softer shadow shapes. I added a NURBS plane and applied an aiShadowMatte so that it would be transparent in the render, but would catch the shadows. I continued tweaking the settings while checking back and forth with the Arnold renderer until I was satisfied with what I could do.
I could name multiple things that I could have done from the start, had I known better. One of my biggest problems was rushing. Even though I was late at every stage of the project, there were definitely times where I should have taken some time to structure my approach. I could've mapped out my original reference picture with perspective lines, which would have given me a plan when I set out to take pictures. With that, I could have also set up the camera and aligned it with the reference image first. That way, I would have been able to check back and forth to make sure everything was lining up with the image. If I was more prepared from the start, I could have yielded a better project, but considering my lack of knowledge, I was probably going to have a lot of problems anyway.
Another big thing that I could've and should've done was receive critique in class. I felt extremely self-conscious about showing how behind I was that it kept me from getting potential help (but it also doesn't help that I get kicked from Zoom every 5 min because my apartment's wifi is bad). Of course, I know I am not the only one who feels self-conscious about showing their work to a bunch of strangers, but I know it is more important to suck it up if I want to grow as an artist.
While I would not consider this an amazing project, I also don't consider it to be my worst. Despite the mistakes I have made, I did the best with what I knew at the time, and I know now what I can do to be more successful in the future.
Comentarios