Martin Thorzen (former CD Projekt Red) talked during GDC 2015 about the process of world creation of The Witcher 3 and the
Martin has a big experience working with artists for a very long time. Although he believes that game artists are capable of creating some amazing things, there are examples, when they fail and use their tools inappropriately. This causes a lot of problem for the technical department and for the game as the whole. In open world games, since they are so incredibly big, this can happen a lot and may slip through quality control more often.
Half of the asset is hidden underground!
CD Projekt Red wanted to get rid of this problems, so they’ve developed a solution for working with assets. Basically they wanted a filter, that can go over 500 000 assets and find problematic ones. There was also an editor that allowed artists to modify assets, which was a nightmare for the tech team.
The developers created a tool called “Database Viewer”. It allowed the team to go through the whole world and all the technical layers of it, all the assets and find the properties they were most interested in: size, materials, lights and other stuff. They’ve messed around with SQL files, created a massive 2GB text file with info and worked a lot with Python. It wasn’t easy, but the service was incredibly useful as it allowed to answer some of the craziest questions the team members had.
You could see the tricky assets on the map, edit them out immediately and find the solution. During the usage of the new database, developers found a lot of interesting stuff, like a shadow-casting light that was 4000 meter in diameter, animated bows under ground, test assets still placed in the level and 105 000 grass entities in an area of 32 sqm. It was a real mess!
Careful approach to assets allowed developers to focus their attention on problems, get rid of all the expensive mistakes and deliver more accurate and logical world design. Too bad the team used Google Sheets to work with all that data. A more streamlined solution like Slemma, would be much more appropriate for messing with such huge amounts of data.