Friday 15 February 2013

The Mortal Engines Project

The last project for Game Production was a character project based on the book Mortal Engines by Philip Reeve. The characters we modelled had to be a self portrait in the style of Mortal Engines. I was excited for this project, I've always enjoyed reading and thought the Mortal Engines story was fun to read with plenty of good descriptive writing of various imagery, especially the characters. I've also always enjoyed reading because it gives me the opportunity to imagine for myself how a story would work visually so this projects has given me the opportunity to develop my ideas into a completed final design with concepts and a 3D model.

As per usual the project had some outlined aims and requirements specific to it. Our models needed to show good appreciation of the final aesthetic, proportion, colour, silhouette and detail density, and the character models were to be judged on anatomical accuracy and artistic interpretation as well as the technical requirements including, an efficient mesh with good surface topology, that demonstrates an eye for sculpting convincing body form, muscles and facial structure. As usual our tutor was looking for an efficient use of textures, conveying the different surface qualities such as skin, clothing etc.

I was a little concerned initially considering the time frame that had been given to complete the project, which included concepting ideas for the character, including sketches of various different ideas considering the different character elements from the book, complete with a final idea and reference drawings from the front, side and back for use when modelling. As many reference images were then collected in relation to my final character idea as well as mood boards and colour palettes. The project then required me to model my high poly version, unwrap and texture it, I then planned on learning some basic Zbrush functions to create and bake normal maps and ambient occlusion maps. My final model then had to be rigged as well as final renders being taken for presentation purposes. A low poly 'LOD' then had to be modelled, unwrapped and textured. Considering all of these components there was an astonishing amount of work to do in the four weeks give in my opinion, and as you'll find out, I did struggle to complete the work that I did.

As mentioned, I started by creating a number of concept sketches and then a final design for my character. In my final design I decided to mix various elements from my four concept sketches. With the 'Mortal Engines' world, I imagined that lots of things would be salvaged to be used like described in the book when London eats another city. My final idea was that instead of my Pirate 'me' character having a plain wooden stump instead of a foot he could have a car piston which could move when walking. This would add interest to my character design and would also create a challenging element when it comes to rigging the character. I also decided to add a mechanical hand on the same side as the foot, the idea is that my character would've sustained an injury to that side of his body and had salvaged parts to repair himself, it also gives the character consistency in the interest of his appearance. The final designs are shown below;



The modelling process went quite well, I feel that I am now at a stage where I am able to model with a reasonable amount of confidence, knowledge and experience in creating accurate models with clean geometry. After first creating a base model, with basic anatomy as a starting point I then began to focus on modelling the head, being as the project was a 'self portrait' I needed to try and get as close a likeness to myself as possible within a reasonable amount of time. I modelled the head using the box modelling method, following the previous method I used from the gladiator project and the tutorial from year one. It went well but slower than expected, and I was having to add lots of loops to increase the geometry meaning I was constantly going over the same area refining and optimising the geometry from all view ports. I finally reached a point that I was happy with, a point were the tweaks I was making were so small it seemed unnecessary to continue for such small details, again, if the time frame had been longer it is certainly an area were I would have continued. I then spent time beginning to add clothes to my body mesh by extruding out from the centre of the body and reattaching at the edges to create the jacket. For the pants I have simply rescaled the legs and added some extra loops where I'd expect there to be more defined creases such as around the knee area and above the boot. I also added some twist to individual loops to make the clothing seem more natural instead of pre-planned. I found that I spent a lot of time modelling the mechanical hand which was not as easy as I'd anticipated, especially in terms of how time consuming it was. Because the mechanical arm/ hand had to be rigged I specifically had to think and plan about which elements to include, where the hand was going to bend and where the joints needed to be placed specifically. Despite this I think the final hand looks great.

Into the texturing process and I was looking at taking a new approach to the process. Usually after unwrapping I will do the diffuse texture first, then go onto the normal map and finally the specular map. This time I planned to use the baking method to bake my high poly models (created in Zbrush) onto the mesh to create my normal maps and this time ambient occlusion maps also. From the baked maps I will then paint the diffuse maps using the UV coordinates as well as the information that has been baked in the the normal and AO maps. The first bake I did was the main body. I decided to do this first as I was trying to use ZBrush to sculpt various elements especially creases in the clothing. As I had never used ZBrush before, and because time for the was short, I decided to stick to the basic tools of ZBrush and use it purely for creating some defined creases in the clothes. ZBrush was a little daunting to get my head around but I managed to develop some simple techniques to get the final effect in the creased clothes I desired. I decided to import the ZBrush model back into 3DS Max to bake the normal and Ao maps in that program instead of using ZBrush's render to texture tool. I was informed by a third year that this would potentially yield better results because I'd be baking straight onto my actual model. This is what I did and I didn't have a chance to experiment with the two options so I cant conclude if one method is in fact better than another. I also baked the individual high poly props including the gun and telescope as individual elements. This just meant that I had to use the mix normal maps feature in nDo2 to combine the various normals I had baked which was an easy enough process. This was the same with the individually baked AO maps but these were just combined easily within Photoshop.

Using as much as my reference material as possible I created my diffuse texture maps. I also tried to texture my model to resemble my initial concept paintings as closely as possible. This meant that for a lot of elements, particularly the clothing I had to work quite hard in manipulating various reference images to get the desired final effect, including merging various colours, patterns, textures, and other things such as stains etc. For the mechanical hand and foot I used some good metal texture reference images I had gathered but when applied to the model it looked quite plain. I decided to also overlay a rust texture which I had taken myself last year which I then painted out in areas to show rust in the desired way, such as in the joints etc. I was quite disappointed, and a little annoyed at how my reference images for my head had come out. They appeared to be very low resolution and not a particularly good quality, even worse when they were stretched to fit my UV's. In the end I realised that my camera had been set to manual focus instead of automatic which was a silly mistake. Fortunately the lower quality reference images seemed to lend themselves well to the model as the skin tone looked much better. The final texture that needed to be done was the specular. This time I decided to try out a colour specular map which was to be placed both in the Specular Level and Specular Colour map slots within 3DS Max. They were quite easy to prepare as from what I can understand it is simply manipulating the diffuse maps to get the desired effect. I did this by changing the hue/ saturation/ lightness adjustments as well as slightly changing the levels when needed. The only major change to colour was in the skin tone. From a tutorial I followed it showed that the skin tone within a specular map should have a darker blue tone to it and this seemed to work well on the models.

Once the model was finished I hadn’t left myself much time at all to complete the rigging. Fortunately I have now rigged a few models, from the gladiator project to the texturing competition last year, and once I had picked up the tools and functions of the skin modifier I was fairly comfortable with getting a basic but solid rig. Unfortunately that was a long time ago and my mind needed refreshing somewhat, particularly with the method of setting up the biped correctly and the specific tools and how to use them properly and efficiently within the skin modifier. Once I had everything set up properly though I began to remember quite quickly and followed a method I used last time for getting a nice clean rig. The method includes attaching verts in the centre of bones a full 1 and verts either side of a joint at a 0.6/0.4 or 0.7/0.3 respectively and then try to follow a steady gradient of weights. I then use the animation I've set up with the biped to change any verts weights independently that look out of place.

With hardly any time left to complete the project I tried my very best to try and optimise my model to a low poly LOD. The aim was to do a 2000 tri low poly model but after many hours refining and adjusting I was able to pull the model down to roughly 2,700 tris which is really still a way off from the initial aim, but, considering the high poly tri count really isn't bad. I also attempted a diffuse bake onto the low poly model, which was a really rushed job, it hasn't come out too bad in all honesty but I know I could do a much better job if I had more time to spend on it.

There are many positives about the final outcome of this project. The main one being that I am extremely happy with the final model as a whole, I think it presents a very interesting character, with a good colour palette which is consistent throughout yet diverse in its presentation.

I am happy with learning some basic uses of ZBrush which has given me a good stepping stone towards better normal map creation and although I’ve already been familiar with baking I think this project has helped develop my skills in this area.

I dont know what more I can say in regards to the time given for this project other than to say I have struggled to get the work done in time. I have tried to balance a quick and efficient work flow whilst still trying to achieve a high standard of work but until you begin the different aspects of the whole process you don’t realise how much time each aspect is capable of consuming. This has been a hard but valuable lesson to learn.

Final renders taken in Marmoset Toolbag
Final renders taken in 3DS Max 2012, showing the final model with normal maps and Ao maps applied.

No comments:

Post a Comment