Film room project, the post mortem

I'll be honest, I'm quite pleased with our final outcome, not necessarily in terms of how much it matches up to the original shot. In and of itself, I feel that our scene definitely holds something of its own, even if we lost a lot of the Pixar magic the original has.

Pixar's Toy Story 3

 

DMU game art students attempt Toy Story

With exception of the characters, who were not part of the brief, it's pretty obvious to anyone comparing the two what the main difference is here. However, its not the case that we couldn't see where the differences were. The problem lay within our technical understanding of a, how to use lighting, and b, our knowledge of how to use Unreal engine. We just couldn't get it to match.

Here's a crude breakdown of the lighting in their scene:

Warm sunlight beams through the window, hits the boxes & floor directly. Warm light bounces from the floor to other areas of the room, lighting it indirectly. Bounced light is picked up by inlays on the doors, the handles on the drawers and other things. Diffused light from outside also enters the room from window behind camera. It gets pretty technical.

Even if we could see what was going on, we just couldn't work out how they did it. We dissected their scene, and it took all kinds of trial and error and faking all kinds of lighting to get as close as we did. There were so many factors that effected the lighting that it usually ended up being a trade off. We would adjust the lighting to make one area of the room match closely, but in turn another area would change completely. This was the same when colour grading the shot.

Throughout the engine part of the project, we felt as though the whole scene was held together by tooth picks, except we weren't sure where the tooth picks were. We encountered a lot of technical difficulties and spent a lot of time looking things up. Things would fall apart for reasons we couldn't find, we'd apply a decal to the wall, and the lighting would break. Life's mysteries!

Knowing what we know now, if we were to do this project again we would probably budget more time towards working within the engine and lighting the scene, as well as take the time to do some learning on lighting. 

On the other side of all this, we considered that Pixar is composed of some of the best artists in the entertainment industry. They would've had a huge budget, more time, many artists and more than likely separate departments dedicated solely to things such as lighting alone. As four students, we think we did an alright job in the end.

Either way, this project was a highly enjoyable and huge learning experience which definitely opened us up to things we can work to improve for future projects. I gained a much better understanding of physically based rendering as well as working in Unreal 4 and had a good experience working within a group.

Film room project, production

In my last post, I talked about our choice of room. Here I am going to talk briefly about how we developed and built the room.

Work on the room began by applying our chosen shot to the Max viewport background, and using perspective match tools to align the viewport grid and begin creating a proxy mesh; a rough 3d block in of the entire scene. This served as a scale reference when creating the actual assets. We each created our own, and as a group decided on the most accurate to work from.

perspective match

Before we dove into any 'final' modelling we were basically tasked with analyzing the scene, and documenting this via studies and/or paint overs. I produced a simple study, focusing on entirely on readability and the colour palette in Andy's room.

As much as I enjoy painting and believe that it is very important to analyze, there were some things about this stage that I personally thought to be unnecessary. This stage was constantly referred to as the 'concept stage' A concept is an idea, and in this project we were already working from a fully realized idea. Andy's room has already been taken through an entire design process from by artists at Pixar, and the concept was already there. Here, we are duplicating the scene as a means to study.

Once the proxy had been made we divided the assets between members of the group and started on the models. We had to really plan our timing as we needed to allow plenty time for texturing and also dealing with technical work in Unreal.

When creating the models, I gathered as many references of the asset in the scene at different angles and used those to draw up some orthographs. I used these as a basis for my model, but eventually deviated away and constantly referred back to the original source material. Despite the chest of drawers being the most complex I actually spent most of my time trying to optimize the boxes, allowing for a good amount of detail without using an excessive amount of tri's.

assets in max

For the floor of Andy's room I used a photo of my own bathroom floor and created a tileable texture from that. The floor was very important to the entire scene, and had to have just the right amount of reflectivity without being too noticeable from a distance.

testing the floor in Unreal

We constantly checked our textures in Unreal, there was a lot of trial and error with creating the maps and seeing how they appeared under different lighting conditions in engine. One of the hardest parts of creating the chest of drawers for example was getting the subtle wear and tear just right.
 

chest of drawers within the final scene

Once everyone had finished their models I imported everything into a single max scene. Even with the proxy as a scale reference certain assets had gone somewhat out of proportion as we worked on them. Using perspective match again I matched each asset to the scene and rescaled everything accordingly before redistributing the resized models which would go on to be used in the final scene.

second proxy

Below is our final shot in engine, with some subtle post process colour grading. I'm going to save my thoughts on the final shot, and in my next post I'll spend some good time talking about what went wrong, what went right and where we could improve.

final shot, Andy's room

Film room project, choosing a room.

Our second project of the year briefed us to recreate as accurately as possible an iconic room from any film or TV series of our choice. Even with a group of four this was an extremely challenging project and had to be completed in three weeks. I can honestly say that in the last four weeks I've already spent more time in the universities labs than I ever did during the whole of last year.

So far, I'm feeling much better about the course and the new course structure this year as opposed to last year. It makes me much happier that there's no longer numerous unrelated and overlapping projects from the different modules. Instead, the visual design and game production modules now seem to interrelate and work seamlessly together. Overall, the course feels a lot more refined and gives you a lot of time for each project. Because of this, I'm personally feeling more motivation to work much harder on each individual project and have it looking the best I can make it, as opposed to juggling time and quality between projects like last year.

To begin our project, we started out by creating mood boards of shots from all of our favourite films or television series with rooms that we'd like to recreate. Here's one of mine-

GoT- The Eyrie

   
At first, we all agreed as a group that the Eyrie's throne room would be an awesome choice. We all love Game of Thrones, and felt pretty excited about the idea of re-creating this room. This was our first mistake. The focus of the project was on colour, lighting and mood. Even though Game of Thrones has a lot of that, we'd basically chosen this room almost entirely on the basis that we enjoyed game of thrones and the fantasy genre, as opposed to considering other factors of the project. In order to progress, we had to distance ourselves from the films/series and any personal attachment we had to them, and consider the individual shots for what they were in terms of colour, light and composition.

Our tutors definitely thought we had chosen something cliché, and encouraged us to choose another film. Max, a member of our team, put together another mood-board with a lot of variety.

Max's moodboard

Gathering other opinions from peers and tutors we found that the shots from Pixar movies were a particularly popular choice. This led to us finding and deciding upon a particular shot from Toy Story 3 that the team all loved, this decision also had positive responses from tutors during presentation.

Moving out, Toy Story 3

What was cool about the final decision to do Andy's room was that even though we all liked the scene, it was a scene we would probably have never concluded upon by ourselves. There was a lot here that appealed to us, particularly the light of the sun hitting the box and floor, and bouncing to other areas of the room. Also, while the scene appears to be relatively simple if compared to some grittier live action films, there is actually a lot of subtle things going on. The reflectivity of the floor, slight wear and tear on furniture and the darker areas of wallpaper where Andy's posters had been and prevented any bleaching from the sun were all things we were going to have to consider when re-creating this shot.

year two, first week roundup

After a long summer spent drawing and painting I've now begun year two on game art design here at uni.  Last year I basically neglected this blog. But this time around its got to be a weekly thing.

Our first project of the year was only a week long and doubled up not only as an introduction to working with PBR (physically based rendering) in the new unreal engine 4, but also as an introduction to working as a team to create assets before exporting them into a scene.

The project was called 'asset swap' and involved creating assets such as a trestle table, bench and a shield for a small viking banquet. It began by assigning one different asset to each person in the group, once that person had developed concepts for the asset, the concept was then given to the next person to begin modelling, and the model passed on again for texturing and so forth. This process of basically meant that everyone in the group worked on every asset at some stage.

Personally I thought this process came with some drawbacks, for instance, when it came to texturing you had to hope that the person before you had made a good job of unwrapping the model efficiently, otherwise you were going to have a harder time texturing. For the most part though, I thought this process helped to achieve a consistent style throughout each asset, as opposed to if people had worked on assets individually.

The hardest part of this project for me was getting my head around the new physically based rendering in UE4. This is a new way of texturing that does away with specular and diffuse maps, replacing them with albedo, metalness and roughness maps.With the old way the amount of reflectivity and shine of a material was defined by the levels of specularity in the specular map, whether this was metal, plastic or skin. Now you can define whether or not a material is a metal, and then you can define the roughness/smoothness of a material, which will dictate how much light information is picked up by that surface. Traditionally you might also paint 'fake' light information such as ambient occlusion on the diffuse texture, with an albedo map however you use purely flat colour information as all the light is defined by other maps and scene lighting. Its an easy concept to understand and actually quite simple in practice, but still something I have yet to grasp properly.