Character Artist and 3D Generalist Jef Bernaers told us about his thesis written at Breda University of Applied Sciences, dedicated to the impact concept art has on character creation.
Hi, my name is Jef Bernaers. I’m a Character Artist and 3D Generalist from Belgium and for the past year, I have researched the impact of concept art on the character creation pipeline. During my research, I have found a large inconsistency between my data and current industry practices.
In this post, I’ll go over why I did this research, how I did it, my results, and how you can implement the results to improve your own workflow.
Why Did I Do This Research?
How much concept art do I actually need? Can I create my work from a rough sketch that I really like, or do I need a finished set of concept art? I think these are questions a lot of character artists have asked themselves before. However, when you look for an answer there is none to be found.
It seems as if people just assume how much concept art is required, but there is no research on how exactly each stage of concept art influences the efficiency of the people who end up using it. Larger companies may have done this research in-house and already know this but have decided not to share their knowledge. Preliminary interviews clearly indicated differences between how different artists perceive the importance of concept art. This is important because if there are different perceptions of the importance of concept art, how do you know which one is the correct one? However, no data, research, or evidence was presented for any of the given reasons. Now, when looking back at the results of my research, I thoroughly believe that the pipelines we are currently teaching have better alternatives, and I’m going to show you why.
How Did I Do It?
I decided to go for a two-step approach. First, I did a quasi-experiment with students from Howest and BUAS to practically test the impact of concept art completion on the quality and efficiency of character artists, followed by a user experience questionnaire. The second step entailed interviews with industry professionals followed by a Q-Sorting exercise.
First, the practical test. The goal of the practical test was to get actual data showing the real impact of concept art rather than just the word of some industry professionals, because how could I be sure that the industry wasn’t just plain wrong? I gathered a bunch of character art students and divided them into five groups. Every group got a different stage of concept art. These stages were: line art, value sketches, color sketches, a character sheet, and a final render, as seen in the following images created by the incredibly talented Alyssa Herman.
By letting different students work on the same concept, for the same amount of time, in different stages of its development, their results could be compared to the amount of concept art they used to create it. In this case, the students worked for two hours, which I describe as the blockout stage of sculpting. After the practical test, the students filled in a user experience questionnaire. The questionnaire allowed for more context behind certain results and possible explanations for outliers.
I followed the practical test with interviews with industry professionals. I had the opportunity to talk to real industry titans such as Charles Zembillas, Nicholas Cole, and Kevin Bayliss. My interviews were divided into two parts. First, I started with semi-structured interviews. They allowed me to learn how industry professionals perceived the problem I raised and showed me what current industry methodologies look like. I followed this up with a Q-Sort. Q-methodology is a research methodology that allows subjective data to be turned into more objective data. The interviewees got images from the practical test and had to rate them on a bell curve-shaped roster (as seen below). The exercise was done four times, each based on different criteria. These criteria were: silhouette, proportions, completion, and concept resemblance. This method allowed me to judge the quality of the results of the practical test. By letting industry experts judge the quality of the test results, I removed my own bias from the data, allowing for much more reliable conclusions.
What Were My Results?
Firstly, this will come as no surprise to anyone, but the more finished concept art is, the more readable said concept art becomes. Data from the practical test clearly shows a steady increase in concept readability for every subsequent stage. I know it makes sense, but now you have proof. You’re welcome.
However, the completion of concept art had no impact on character artists’ quality or efficiency. The following graph shows how students rated their concepts’ readability (black), their efficiency (grey), and how industry professionals rated the quality of the students’ results (orange). Industry professionals were consistently not able to distinguish sculpts created with a high amount of concept art from those created with a low amount. If industry experts are not able to tell the difference between sculpts created with a final render from those created with lineart, then why would you wait for the final render to start sculpting?
Lastly, interviews indicated a difference in priorities depending on the size of your team and whether you practice outsourcing. Smaller companies could potentially get away with less detailed concept art because the team is closer, and communication between the team members is smoother and easier. Less rigid concept art could allow for more creative freedom for the artist and more creative freedom could potentially improve job satisfaction and therefore increase productivity. Larger companies, where departments are more distanced or even outsourced, could benefit from more defined, rigid concept art. This ensures a unified vision across different departments and teams. It also ensures artists don’t go in a different artistic direction than required, a risk that is potentially higher when outsourcing.
While my research is of a small scale, it might have bigger implications for our industry. I effectively researched about 5% of the character creation pipeline and even there I have found a large inconsistency between my results and standardized industry methods. This indicates that there might be many other parts of the pipeline that can be optimized.
Currently, when you search the internet for a character creation pipeline, you get what could be called a “waterfall pipeline”. Every department is kept separate, and every step only begins when the previous step has been completed. This is similar to the Waterfall methodology, which game developers have consistently been replacing with Agile methods like Scrum over the past years. If Agile works on large-scale game project management, maybe we can also apply it to the individual parts of the projects. Going for a more iterative workflow, with less heavily guarded borders between the individual steps, might also prove beneficial to the individual pipelines of a project.
So, I suggest the following. In the following images, the image on the left is what could be described as a basic 3D character creation pipeline, much like what you would find if you google for one. On the right, you have what I would change based on my results. Blur the lines between your departments, don’t keep your artists separated. Rather than letting your concept artist(s) go at it alone, involve the character artist. Once a concept is chosen and a lineart has been defined, the character artist can create a blockout. This can help the concept artist get a better view of how their ideas translate to 3D and at the same time it might give the character artist some extra creative freedom that could even inspire some new ideas for the concept artist. They are all creative individuals, and promoting more cooperation can only prove beneficial to the quality of the work they deliver. Additionally, when you do get to the sculpting stage, you get to start with a blockout that has already been reviewed by the concept artist. Therefore, you decrease the chance of having to do any more major changes to the design of the character.
I have only changed the start of the pipeline because that is the only part I researched. But there could be many other improvements to be made in the other stages as well.
This is only a recommendation of course and it will not be applicable to all companies or projects, because it is impossible to create such a universal pipeline. Therefore, I strongly recommend any company critically reflect on the reasoning behind their pipeline decisions. Do you actually know if you have a good pipeline? Did you test its efficiency? Do you have any data behind your decisions? Or did you just assume it was good because it worked? Staying critical will always be important if we want to keep making better games and better work environments for the amazing people that create them.