Tuesday, May 21, 2013
So I went back to this lady and tried to finalize the model. She's made up of multiple parts so I tried to move the edges as close to each other as possible. You can still see the seams but hopefully after painting, you won't. I added in some edges where her gloves end. I added a few folds near her crotch and elbow to show her suit bending. I should have added in more but they weren't coming out right. I spent too much time on the butt again but finally got it looking better. I tried to smooth out different areas so it looks like she's wearing something and is not naked.
I don't like that's she's bald. After I'm done with the book, I might go back and add some hair.
Back to the book. This chapter was on using 3D scanned data. Above I have the results of a scanned head. You can see there are many problems with it and I am going to have to fix them. The book is a few years old so scanning technology might have changed since then. Problem areas for scanning are hair which is why she's wearing a cap, shiny things like the eyes, undercut areas like the nostrils and ears, and horizontal surfaced like the top of the head and chin (you can see the hole there).
Another problem is that the scan is made up of triangles which is a big no no. It has to be made up of quadrilaterals instead. The first step is to delete the big spikes shooting out of the head in a program like Maya.
Then I can bring that back to Mudbox. Next I add in a default head model which is yellow here. Then I use some sculpting tools to shape the head so it matches the scan as much as possible.
After that, I extract a displacement map of the scanned head. Now I can take that displacement map and place it onto the default head. The result is on the above left. It looks exactly like the scanned head and you can see that all the holes are now closed up. We don't need the scanned head anymore and will work with just the other head. If I turn on the wireframe, you can see that it's made up of quadrilaterals instead of triangles. Then it's just a matter of smoothing things out and sculpting where needed. I probably over smoothed below. I started to work on the ear when I remembered I just sculpted a head so I stopped here.
Now for something completely different. I took a break from the book like I usually do. A modeler doesn't always model using millions of polygons. Depending on the application, like a video game, they have to use much less. So I had an idea of doing a project with three models; one with a very low number of polygons, one higher, and one with millions. The models had to be related so my mind drifted to Pokemon and how they have three evolution states. I wanted to do Charmander since he was the Pokemon I picked in the videogame way back in the day. Unfortunately, his evolution stages didn't differ enough for me so I settled with Squirtle. So Squirtle was going to have the least polygons while Blastoise would have the most. The first step to find reference images. Despite the number of picture of him online, I could not find a profile shot or a good head on shot so I just used the above image as a reference. Then I went back to Maya, for the first time in a long time, and produced the model above. It wasn't super easy because there's a balance trying to make it look like something while using the least amount of polygons as possible. I think I ended up with a 300 polygon model.
After that, I unfolded the model to make the map. Then I imported it into Mudbox for painting. Painting him ended up being much trickier than I anticipated. When painting something realistic, I'm working with lots of colors and things can be blended together and imperfections make it look more realistic. Here I'm dealing with only with four colors so everything has to be nice and crisp and perfect. Also the model is very boxy. So when I was drawing the black lines, it looked nice and smooth from one direction but bent and crooked from another.
For the face, I just pasted in one I found online but then I had to retrace all the black lines to make them sharp. With the painting done, I used Mudbox's posing tools to make him look more natural. I just moved one arm up and then curved the tail so it wasn't pointing straight back. Below you can see what my map looked like after it has been colored. Overall he turned out ok. The head shape could be better. I might fix that later.
Next I moved onto the head. The book let me do my own thing for it. I didn't want to sculpt a face from scratch so I decided to find some reference images. I picked an actress that I liked. I won't tell you who it is because my finished product does not likely resemble the person. One nice thing about picking a famous person is that there's plenty of photos of them online. I couldn't find a nice profile picture though.
Trying to sculpt a real life person is tough so I was getting frustrated at times when things weren't looking right. I think that's one reason I didn't save any in progress images since I was never satisfied with what I was doing until the later stages. While the final result may not look like the person, I did end up with a fairly good looking face. The actress is young so I didn't have to worry about any wrinkles or anything like that. The face might be a little too smooth and perfect though. I skipped the fold in the eyelid but maybe I'll be able to paint that in later.
I figure I should say something about the process. The ear I sculpted from scratch. I didn't try to make it look like the actual person's ear. The nose was tricky to get right. It wasn't hard to make it look like a nose at different angles but when I looked straight on, I could see that something was off. Since I had a straight on photo, I knew what it should look like but it took some time to get there. I'm referring to the nostril shape and the bottom of the nose area - around the nostrils and how it connects to the area above the mouth. Unfortunately there weren't many up the nostril shots of this actress. For the eyes, I created a sphere. I had to make sure it was the right size and in the right place. Because the eyelid has to be sculpted in relation to the sphere, if the sphere is wrong, then the shape of the eye is thrown off. The mouth was tough to get started mainly because you're working with a single sheet of polygons when in reality the mouth is two lumps of flesh touching each other. But once you get the general shape, it becomes easier.
Continuing on with my book, I'm back to the lady I began awhile ago. Here is the model I was provided with. The author of the book decided to chop the body up into multiple pieces; head, hands, upper and lower body. The main reason being that the head and hands require more detail so I can increase the polygon count on those areas without having to do it to the whole body. The problem was that the book never addressed what I should do in the areas where the parts meet. The hands and neck aren't too bad but there's a huge seam running across the chest. Even more annoying, in the book's video, the author just had a single body piece and a head and said it was just easier working with two pieces. Then why didn't you let me do that too, you bastard?
Anyway, here's the result of me working on the body for the first time. She looks kind of creepy with the unsculpted face. This character is supposed to be a female android type thing from the future. She'll be wearing as skin tight body suit and be bald. This conveniently means I don't have to sculpt clothes or hair. I spent the most time on the breasts and butt because I am a perv. No, I mean I had a reason. It was because they had to be symmetrical and also had to have the right curvature. Having a nice curved surface is sometimes tough in Mudbox when you're making multiple strokes on top of each other. There is a smoothing tool but it tends to flatten things instead of maintaining the curve. After roughing out the body, I moved onto the hands. Hands are tedious to sculpt due to all the bumps and depressions on so many fingers. Luckily, she's gloved so at least I didn't have to make nails.
Monday, May 13, 2013
I used an image I already had to paint on the eyes quickly. I also added some pink along the edges. Next was the hair. I was a little worried about how I was going to do this. Luckily I found a good combination of stamps and settings that allowed me to easily do the job with much better results than I expected.
Lastly, I picked out a good skin tone and painted that all over while adding some pink highlights here and there.
Below you'll find the finished color map.
After coloring, I added in some bump maps. I just just used copies of the color map for the sweater and hair as a quick way to make the bump map. It's fairly subtle so I turned off the coloring so you can see what it actually looks like. I also added in a specular map make the shine of the eyes and highlights in the hair.
Now let the painting begin. I started with his pants. I painted a white base layer with blue on top of it. Then I erased some blue to get the worn look on his knees and added in some dirt. I spent too much time on the pants though because when I was done, I zoomed out to see how it looked and realized he has tiny legs and they're barely noticeable. Oh well. I moved onto his shoes making sure to add dirt and scuffs to them.
I moved on to the slingshot. The rubber bands and pouch just had a single layer of color. The wood had a pale yellow base coat. Dark brown was placed over it and fine lines of yellow painted on top of that. Then the top ends were erased to reveal the yellow below.
The sweater was a little more tricky. I had a base layer of dark green and then used a sweater stencil to paint on the lighter green pattern. The hard part was that the sweater is wider on the bottom than the top. I could have made the treading get narrower as it went up but I thought that wouldn't look right. Instead the treading is the same size all over but that meant that there would be some visible seams where one painted section met another. I tried to minimize that in the front by painting a larger continuous portion there. You can still see the seams but I don't think it's that bad.
With two Mudbox books completed, I wanted to do a side project. I wanted to take the character that Shiu Pei drew that I modeled awhile ago and color it using Mudbox. First, I had to take my original model and unfold it to create the map. This took longer than I expected. When I made him, I was just focused on the end result and didn't model him that efficiently since I wasn't expecting to work on him again. So I had go back into Maya and clean up the model as much as I could. Some parts of him had way more polygons than were needed, some surfaces were reversed, etc. After that, I imported him into the unfolding program that one my Mudbox books showed me how to use which you can see above.
The next part took awhile because I had to relearn how to use the program. While the program is really good, it's not super clear how everything works. There were a lot of hotkeys and such I had to look up. But eventually, I got the resulting map above after cutting up pieces and flattening them out.
With that done, I could finally import the model into Mudbox. The goal was to just paint him but since all the tools were right there, I decided to sculpt a little and make him look better. I gave him some knees. There were some gaps in his hair so it was easy to just use the grab tool to close those up. The shoes looked fine from the front but were actually fairly misshaped which you could see from different angles so I fixed that. I spent awhile on the face. His head was very cube shaped so I tried to round it out a little more. I adjusted his nose so it wasn't a flat blob and fixed up the mouth. While I was doing this, I didn't think I wasn't making that much of a change but when put side by side with the original, his face looks much better.
Wednesday, May 8, 2013
Another filter Mudbox has is showing the normals of the model which can be used to make a normal map. Normals are the perpendicularness of a surface. Normal maps and bump maps are similar in that they give an illusion of three dimensionality. Bump maps simulate just depth (up and down) so they can be black and white images. Normal maps contain the three dimensional direction of how light bounces off the surface so they need to be color images to store more information. Above is the normal map of the rock I made earlier.
I can save the map and apply to a flat square and now it appears as if there is a three dimensional rock there. If I tilt the viewing angle (below), it changes how the light bounces off the surface as if a rock was really there. However if I keep tilting, you can see how the square was flat all along. In a way, normal maps are like holograms. Holograms change the way light bounces off of it when you look at it from different angles to create a three dimensional effect despite being a flat object.
With that, I'm done with this book. I went through it fairly quickly. They didn't make the most interesting blog posts though. I thought the book was just ok. I wasn't a fan of the format of having a bunch of small lessons. I did learn and do some new things but I'm not sure how much I would use that in the future. Anyway, I promise there are more interesting blog posts to come.
Now I'm going to make a rock. Very interesting stuff. I took a sphere and roughed out a rock shape. I found a picture of a rock and used that to paint the colors onto my model. There are different kinds of maps I can make in Mudbox, some of which you've already seen like color map, specular (shininess) map and bump map. As a short cut, I can copy the color map and convert to different map. So the bump map creates the illusion of depth depending on the value of black to white. By duplicating the color map and assigning that to the bump map as well, all the dark areas appear to be recessed in the rock.
I can do the same for the specular map. Now all the white areas automatically appear more shiny than the dark areas which you can see on the right. On the left are just random dots of shininess I made to represent the crystalline structure of the rock. I can have multiple maps so both can be layered on top of each other.
Above I have the final rock. I sculpted in some more cracks and crevices to help hide some problems in the initial paint job where you could see some seams in between paint strokes.
Here I have a problem that can come up when using stencils. I want the word to appear on the sheet, however the sheet is crumpled so the word isn't going to come out correctly.
One thing I can do in Mudbox is flatten any 3-D model into its 2-D map and then paint directly on that. This is helpful to check if I missed painting any spots since some parts of the model could get blocked by other parts and when it's flat I can now see everything. It's also useful in this case. Here the sheet's map ends up being just a simple square. Now I can use the stencil to paint the word and when I unflatten the model, I get the result below.
Instead of switching back to my other Mudbox book, I decided start on a third one. The main reason was that this new book was written for Mudbox 2013. The book I just finished was for 2010 and the other one for 2011 so I just wanted to see what 2013 had to offer (there aren't too many Mudbox books out there). This new book is a little different. Instead of long lessons, it basically has lots of tips and mini-lessons. There's not a lot of visually interesting things to show and some things I just read instead of following along. So I went through the book fairly quickly. Anyway, above is supposed to be a stone tile I made starting out with a basic cube and using just a single stamp to sculpt.
In case you forgot, stamps are sort of like the brush tip I can use to sculpt with. I then went over making some stamps in Photoshop. They just have to be black and white images where white indicates where to push down and black is where nothing happens. Along with just drawing one in Photoshop, I can make a stamp out of something I sculpted. For example if I needed to make a wall full of bullet holes, I can sculpt one hole, make a stamp out of it and then use the stamp to make a lot of copies. So on the left, I made a hole and on the right is a filter that Mudbox has that shows depth in black and white. I can export that image to Photoshop, invert the colors and import it back as a stamp.
Now I can use the new stamp to make copies of the hole. Below I spent some time tweaking the material of a model. I can adjust color, the reflected color, shininess, etc to make gold on the left and glass on the right.
Next was a lesson in extracting maps from the sculpted model. I've done this before so the basic idea is that there is too much information in the model and some applications like a video game wouldn't be able to handle it. One way to sidestep that is to extract information from the high rez model and apply it to a lower rez version. Above we have a displacement map that Mudbox generated off of my model.
Now in Maya, I can take that map and apply it to the low rez version on the left so when I render the model, it'll look like what's on the right. The displacement map is actually changing the shape of the model at render time so it still takes a fair amount of computational power. An alternative is to use a normal map which Mudbox can also generate.
A normal map tells the computer how light should bounce off the model and isn't that graphics intensive. So when I apply the normal map, I can see the results in real time as opposed to just at render. Above I have the normal map applied and the green lines represent the actual polygons in the model. By using the map, I can get the look of a 10 million polygon model on one that's just 5000 polygons.
So with that I'm done with the book. I decided to try out one last thing. I went back to the girl head I sculpted and used the reference photos I was provided to try to paint the real face onto the model. Since I had a number of photos from different angles, it was able to get decent coverage and not end up with obvious seams where one photo ends and another begins. I didn't try to get it perfect or finish though since my sculpted hair was different than the photos, but I may come back to this in the future. Otherwise, I thought this was a good book. It wasn't super in depth but I liked how it went over anatomy. It seemed like the book was more geared towards real life sculptors who wanted to learn digital sculpting.
With the model done, it's time to paint the guy. I used a skin color and covered the entire body with it.
The book provided me with a texture to apply to the skin as a bump map. The texture wasn't quite human looking. The book admitted to this but since the book's model was more monstery it was ok. I just toned down the strength to lean it back towards being more human.
After that, I lightly applied red and blue to various parts of the body to represent the blood and veins beneath the skin. It's fairly subtle but does break up the flat skin tone a little.
After doing the subtle coloring, the book told me to ruin it by putting dirt all over him. So that's what I did using a picture of dirt as a stencil. The book also provided a stencil for the eyes. My model has less squinty eyes than the book's so the yellow on black looks a little off to me. Since I was making a muscular guy, I named my file "juicer". Ironically, the book had me apply a tattoo that a juicehead from Jersey would have. Lastly I changed the material of the model so he would be less shiny.