“Smile” short film case study Part 1, Modelling and Texturing.

Hello and welcome to a brief case study about the short film I made with a team-mate (Philip Zarcilla) for my final year project at university. This is part 1, which shall cover modelling and texturing. I will go through some of the different styles of modelling used on the project as well as having a look at some of the designs used, and then a look at the texturing. You can see our finished film here.

Character Design And Modelling

I will begin with a brief look at some of the concept art. There were many many different designs for the main character. It took us quite a while before we were happy with it. We were looking for something that looked androgynous, human like and could emote with only its eyes and body language. The story is set in the not too distant future so it wasn’t desirable that the robot be too advanced, we had a lot of cogs here and there on the robot. This allowed us to give it some interesting secondary animation which could give the viewer a hint on its emotional state and it also gave it an interesting “Victorian contraption” look. As it was being modelled there were several changes and tweaks made to areas around the joints. This was because parts would intersect each other as the limbs would bend which was not what we wanted.

The design was very complicated. There were many separate pieces and all together it was made up of around two hundred thousand polygons. It would be subdivided several times at render time to make it smooth.

You can see a textured and posed version of the final design for the robot character here.

The most difficult things to model were the outer shells. It can be very tricky to make grill holes in hard surfaces without getting some sort of undesirable look to its corners or edges surrounding the holes.

Character Texturing

Here is a look at some of the textures used on the robot. about 90% of the texture maps were made by my team-mate, below is the colour map for the head and limb frames. Most of the maps were 2048×2048 in size.

Designing And Modelling The Props For The Library

While we were designing the main character we were also designing the environments and props. There were so many different props that needed to be made. The idea was to make all the environments look like they were suddenly abandoned with everything left in place. The library environment had the most objects in it and all of them had to be designed. Here is an example of one of the design sheets created for the props.

Most of the props were made to be turned into subdivision surfaces at render time. This meant that I needed to use specific techniques in order to make it look right when its smoothed. This means having support edges near areas where I wanted sharp corners, just bevelling the edges was not good enough. After a prop was modelled it was then UV’d and then sent off to my team-mate for texturing, Then I would set up shaders and apply the textures to them.

Books

Because half of the film takes place in a library, the most important props were the books. A library is only really recognised as a library if it is full of books. This became a problem because we couldn’t make thousands of individual books. What we ended up doing was model 33 books and then duplicate them many times and shelve them in different configurations.

All the different book types had very similar UV layouts. This made it easy to make a lot of textures because the templates were almost all the same. Once all the books were made I then had to lay them all out on the shelves. This was very tedious, there was a lot of duplicating, moving and rotating. The books were arranged depending on the shot. If the camera was only going to see one bookcase or half a bookcase then I would only need to lay out enough books to cover the bookcases in the shot. In some of the wider shots there were several thousand books in the scene.

The floorboards were created using displacement and normal maps. This was more for experimentation, I wanted to see how far I could push the displacement capabilities while at the same time not use up too much memory. I was pretty happy with the results but if I were to do it again I would model the floorboards individually using basic box primitives with bevelled edges and just apply normal maps.

The Factory Environment

The factory environment didn’t require as many props as the library, which was good because time was running out and we couldn’t afford to make a lot of props to fill it up with. The models were very simply modelled, most if not all the models did not have many polygons. Luckily most of the objects in the factory were hard surfaced metal objects so there was no need to go overboard.

There was also a few dead robots lying around the factory. These were easier to design the the main character because they didn’t have to do anything or be on screen very long. They also needed to be very simple in design since they were to be inferior robots.

I learn’t a huge amount on this project, the best way to really learn how to do things is to dive in at the deep end and just do it. I was modelling things every day and after a little while it just became second nature and I got alot done in quite a short time.

So there it is, a brief look at some of the modelling techniques and textures used on my film. Part 02 will be a look at the rigging of the main character.

Sharkface Version 2 WIP

I decided to have another crack at my “Sharkface” character now that I have learnt more about 3D sculpting applications like Mudbox.  It looks very different and less rushed than the original version and I’m really happy with how it has turned out so far in comparison.  I probably only need to do a few more hours work to it and then it should be done.  I will be adding things like high frequency skin details, wrinkles, skin pores.  Then the next step will be to start making texture maps and start work on the shaders.  I plan to use displacement maps on a relativly low poly version of this model to mimic what would be done in a production environment.  As always I like to include turntable movies enjoy that and some pics

 Turntable Movie

 

 

Textured Rodent

Here is a quick update on my rodent project.  It has been textured and shaded.  I mainly used Photoshop for texturing and bodypaint3d for painting over any seems.  I used Maya’s Skin shader to get a realistic subserface skin look to it.  I am pretty happy with it at the moment, I think with proper lighting in a scene it will look 10 times better.  The next step now is to begin rigging it, which is going to be very tricky but I like a good challenge.

I used a 2K map for the body and tail, the ears used 1K maps because they are small and its silly to waste memory on something that isn’t very big on screen.  I also used both displacement and normal maps, it’s good to have both because on most models Its difficult to be able to subdivide a model enough times at render time without it eating up all the memory so I subdivide it a few times and add displacement, this gives it enough detail for a fairly good surface read and then I add the normal map to add the rest of the detail.

You can view a turntable movie of the finished rodent model with textures below:

Finished Hairless Rodent Sculpt

I finally finished the high resolution sculpt of my hairless rodent model.  It was a very tricky model to sculpt, skin wrinkles can be very difficult to make look good.  I did a lot of research in cloth folds to help me get an idea of how wrinkles and folds look and where they go.  Most of the time I spent on this model was refining techniques  for creating skin folds and wrinkles.  I think I’ll make a tutorial on skin fold and wrinkle sculpting techniques in the near future.  It has been the most difficult thing to sculpt so far using mudbox and I’m quite happy with how it turned out though I think it could do with some major improvements in certain areas but I want to move on and begin texturing, shading, rigging it etc.

Turntable Movie

The final high resolution model has just over 10 million polygons. the next step for this is to extract normal and displacement maps, then get on with creating textures and shaders.

UV Mapping Techniques and Workflow.

This tutorial is designed to show most of the UV mapping techniques and tricks that I use day in and day out. This is more of a brief case study rather than a step by step tutorial but it will give you, the viewer, an idea of the work flow I use when UV mapping something tricky like a human skull. I hope you enjoy it and learn something.

We will begin with a look at the model itself. Its always good to unwrap Uvs on an object that is not to detailed because it can get difficult to edit quickly and I am going to be using UVlayout which automatically smooths and unwraps a models UVs and it can get very slow if there are thousands of polygons to unwrap.   The model is split into three parts, I decided not to have the mandible as a separate object from the rest of the skull because it can be easily split up in UVlayout. The teeth though are separate because they will make a lot of UV shells and its best to have them separate from everything else.

In figure two it shows the model being imported into UVlayout, in the “load options” the “new” box is checked, this clears any previous Uvs that are written into the .obj file. If the “edit” box is checked it will import any Uv’s written into the file so you can edit them.

The next thing to do if the object is symmetrical is to tell Uvlayout to find its symmetry, this helps a lot because it mirrors over any cuts you make to one side and when it unwraps the shells it makes both sides match. This can be seen in figure 3, also I have split up the mandible and main skull object.

In figure 4 I have spent a little time cutting up the model into chunks that can be easily unwrapped, the trickiest area was around the side of the skull. It can sometimes take a while to get these things right, there was a lot of back and forth with cutting up parts and seeing if it unwraps well, and if it didnt unwrap well then I made adjustments and unwrapped again etc.

In figure 5, it shows the process UVlayout goes through as it unwraps the UV shell. It can sometimes take a few minutes depending on the polygon count of the object. you can see I let it calculate a little too long and there are overlapping UVs near the face of the skull, I fix this later. Also notice that with symmetry on the UVs are identical on each side.

It didn’t take too long to unwrap the rest of the shells. You can see in figure 6 in some of the UVs there are some areas of blue or red, this means that areas with red aren’t getting an equal amount of UV space in comparison with all the others UVs, and blue means the polygons are getting too much UV space and green of course means the polygons are getting the right amount of UV space.

In figure 7 I have imported the Uvs back into Maya and applied a “UV display” texture to it. This allows me to easily see on the model if there is any pinching of the UVs or any areas of weirdness.

There are some areas that need adjusting on the mandible. I mainly use UVlayout as a tool to make UV shells, I do all my relaxing and adjusting back in Maya. This is shown in figures 8-10. My technique for relaxing UVs in Maya is to use its “Unfold UVs” tool. I found the best way to use this tool is too use its “unfold constraints”. This means when I relax the UVs it only allows them to move either vertically or horizontally, when I use this tool without “unfold constraints” I sometimes get undesired effects. I unfold vertically and then unfold horizontally.

As you can see in Figure 10 the “unfold UVs” tool has performed perfectly at eliminating any “UV weirdness” on the Mandible.

The next place to fix are the orbital sockets shown in figure 11. Areas like this can be very tricky to unwrap if only a single planar projection is used. I found the best way to improve upon this area was to use two planar projections on the Y axis.

Figure 12 shows what I mean, It effectively turns UVing a bowl shape into Uving something less rounded which will get less distortions. Though for some shapes its just not possible to get an absolutely perfect UV space distribution without a lot of seams.

As you can see in figure 13, I welded together part of the seams on the orbital sockets, this makes it a little easier to texture if only photoshop is being used but luckily these days 3d projection painting in applications like bodypaint are used to paint over seams.  I also like to make the UVs resemble the model as much as possible, this makes it so much easier to make textures for it.

As you can see the UVs for the orbital sockets have been improved quite a bit. Almost done now with the general editing and relaxing of all the shells, I went through to all of them and used the “unfold UVs” tool. I needed to quickly put it back into UVlayout and edit the main skull UV shell with the overlapping parts.

It normally doesn’t take me long to make these adjustments, After a lot of practice at Uving things I get an idea in my head of what probably needs to be done to fix any issues, there isn’t really a huge variation of UV problems and if you can fix one thing its pretty easy to fix another.

Now the final step is to arrange all the UV shells into the UV grid so they can be exported and texturing can begin. Maya has a fantastic tool called the “layout” tool. This takes the UVs of an object or many objects and arranges them in a tidy way. It also scales all the shells in relation to each other. This is important because it means the UVs for the orbital sockets wont get more UV space than the mandible, everything gets an equal amount of UV space for its size in the “3D world“.

The layout tool did a great job of scaling everything and making it fit within the UV grid but doesn’t always arrange the shells in a good way for texturing. I always go in afterwards and arrange them the way I want, this makes texturing easier if everything is arranged in a logical manner. I also added the teeth UVs to this as well. They were very simple to unwrap, I simply put them into Uvlayout all as one object and then split them up. I gave them quite a bit of UV space for their size in relation to the rest of the model, the teeth is something that people like to make detailed so its good to provide them with enough UV space.

That’s pretty much it. I hope you learned something from this, more tutorials are coming.

“Worm fish”

“Worm fish” is a concept sculpt that I did so I could refine my digital sculpting techniques. I haven’t had the opportunity to really see how far I could go with high resolution sculpting applications such as mudbox so I decided to throw myself in at the deep end and see what I could do if I began with the simplest of base meshes.

I have seen a lot of demonstrations on the Internet where someone takes something as simple as a cube and turns it into a monster or a highly detailed human head etc, I decided to try it myself and see if I could make something good.

After a short time of using mudbox’s “grab” tool to block out the basic shape I then subdivided it and began making the tale thinner towards the end and make a hole where the mouth was to be. From their on I sculpted the rest of the forms that made up the main body shape, as I subdivided the model more and more I used a smaller brush size and started sculpting the smaller details such as wrinkles and the protruding vertebrae. Once I had got the model to its final stages I exported the lowest level back to maya so I could give it a basic pose. I then exported that mesh back into mudbox as a layer effectively creating a blendshape in mudbox. That layer was then flattened like when using photoshop’s layers and I adjusted the mesh according to its new pose. This included adding bigger wrinkles and bulging the muscles around areas where there were bends.

The mouth was by far the most complicated area to sculpt, There level of detail in the mouth required the model to be subdivided to a very high level. The final polygon count was just over 19 million polygons, this meant the mudbox file was around 2gb is size. I wanted to take advantage of zbrush’s turntable rendering tools and its matcap materials to bring out all of its details but Zbrush was unable to import a model with so many polygons in a single mesh. I needed to cut the model up into 5 parts, each with about 3.8 million polygons and then import them as subtools.

For a turntable render, click here

Rodent Study

Before I began work on the rodent project I needed to do a lot of research into rodents.  I gathered up as much photographic reference and anatomy charts as I could on the subject of hairless mice and rats and began making sketches. I did sketches so I have an understanding of the anatomy and dimensions of the rodents. This helps because it gives me a mental blueprint of what I’m modelling and allows me to progress quickly. Its almost the same as having a physical sculpture in front of me to look at.

I also found this study to be particularly helpful for modelling “wrinkle head”. Hairless mice and rats have very wrinkly bodies. They look very much like hairless cats. It can be seen in some of my sketches of hairless mice that there is some wrinkle detail drawn in. Wrinkles are tricky to draw and model and I found after this study that I have a much better understanding of how to make wrinkles and how to put them in the right places.

“Wrinkle Head”

This is a small task I set myself so I can learn more digital sculpting techniques. The design is of something I found on the website of Carlos Huante, who is an amazing concept artist for the film industry. The particular design I’m working from is good because there is alot of cool details like wrinkles and muscle creases etc and I got to experiment with ways of detailing teeth and gums.

Turntable Movie

Even though it was sculpted in Mudbox I found that for presentation purposes it was better to import the model into Zbrush to render turntables and take screen shots. Zbrush’s “matcap” materials are good for showing off the details while also having the model antialiased.

The Basemesh was modelled in Maya and the UV’s were unwrapped in UVlayout and Maya. Because the poly count is low and the topology is very simple the UV’s were unwrapped without much trouble at all.

It’s very important when you are creating a base mesh for high resolution sculpting, that the polygons are equally distributed all over the model. Otherwise when the model is subdivided there will be areas that have a very dense distribution of polygons and other areas that arent getting enough polygons.

This is bad because it means more subdividing is needed in order to get enough detail in all areas of the model, which means the polygon count and file size sky rocket very quickly and it becomes difficult to manage.

Old work from early 2008

I thought I might aswell post up some of my old work from my old site.

This is a character model that I made for my First showreel.  If you are familiar with the “metal gear solid” video game series you might recognise him as “Revolver Ocelot”.  Everything was modelled in maya and tweaked in zbrush.  For reference used google and 3d.sk to get a good idea of cloth folds and generally how clothing looks when its on a person.  I also bought a small Action figure of The character and used that for reference too.  Its alot easier when you have something infront of you to model from, this is because you can see it in three dimensions and see it from every angle.

When I modelled the gun for this character It took me about a week because I couldnt find enough reference pictures of a Colt Single action army from all angles.  I like to try and make my models as accurate as possible and not having enough pictures from all angles slowed me down.  I think I did a good job of making it, Hard surface modelling is not has difficult as organic modelling but good topology and an economic use of polygons is still very important.

Here is another piece from my old showreel.  The main inspiration came from the clown from the “spawn” comics.  I quite enjoyed the challenge of modelling someone overweight.  I created a very basic basemesh in maya and then did all the detailing in Zbrush.  I wasn’t very experienced with Zbrush at the time so I found it tricky to create certain shapes and I had to discover techniques on making certain details.  as I look back at it now I think of it as a good attempt of what I was trying to convey but it doesn’t really jump off the page as being anything special.

This was one of the first head models that I made in mudbox.  I used the same basemesh for this head that was used for my “Revolver Ocelot” model.  It took me about 12 hours over a few days to finish.  It took me so long because I still wasn’t very experienced with digital sculpting applications like mudbox and zbrush.