Interview with Hiroshi Sakakibara, Environment Artist @CDPR [Translated]

+
Good interview composed of three sessions. I translated the most parts.

Episode 1

・・・​

▶︎ A brief history of Hiroshi Sakakibara currently working for CDPR as an Environment Artist: (1:06-)

  • Majored in Western history, Kyoto University, Japan; His research theme was "Trade in the Indian Ocean.
  • Graduate school research on Africa. Conducted field research on architectural decoration in Tanzania.
  • Began working as a CG artist for a game company (2K Czech) in the Czech Republic. (2012-)
  • After living in the US for a while (for transferring to Hangar13), moved to CD PROJEKT in Poland (2017-)
Q: Why did you decide to go into the game industry when you studied Western history at university? (1:11-)
I've always liked drawing pictures and building model kits since I was a child. At the same time, I was also interested in history, so I went into that field at university, but while I was still in uni, came to think about doing actual fieldwork, not just historical research, so I decided to go to Africa during graduate school. The destination was Zanzibar Island in Tanzania, a trading city crowded with people from Africa, India, and the Arab world. It was a city where wealthy merchants had a tradition of decorating the front doors of their homes with gorgeous wood carvings, and many of these decorations still exist today. In order to study the local architectural decorations and find out how the locals made their designs, I apprenticed myself to local craftsmen and worked with them to actually make doors on my own with chisels. As I did this, I began to realize that I preferred to make things myself.
Also, one of my friends from uni was a big FF14 fan, and when I watched him play, I had a feeling that game environment design was so amazing, which was connected to my experience in Tanzania, and I finally came to the conclusion that making game background would be the best for me. After that, I decided to start studying not only CG but also painting again, which was always my favorite since I was a kid. Then, I put together a portfolio of my work and sent it to a game company, which led to my being hired.

gamesampo02.png

Q. Is it possible to get hired even if you have no experience in the game industry at all? (4:24-)
That was not the case for me, and of course it will be different for companies that need people who can work immediately. One thing I can say is that some kind of ability like “drawing pictures” is one of the highly valued skills in the CG industry because even if you learn the software to make CG, it is meaningless unless you know what kind of pictures are cool. If a person who wants to go into the CG industry in the future has a special skill such as "drawing cool pictures" or "making cool models", then it can be said that (s)he already has enough qualities to enter the CG industry.

Q. What kind of study do I need to do to become a CG artist? (5:16-)
It doesn't matter if it's painting or 3D. I think it is important to have the skills to create something ”cool” by yourself, whether it is pictures, models, or anything else. Just learning CG software is not good enough; there must be a lot of people who are hired without it. At least in my case, the drawing skill helped me in my career.

Note: If you want to know more about Mr. Sakakibara, his twitter is here.
Also, I recommend this page, which is in Japanese ofc, but I think Google translation is good enough to understand.


・・・​

▶︎ Afterlife(7:13-


Since the Afterlife was a converted mortuary to bar, the morgue tables are still in use, and the many grooves on the floor are probably to drain something (like blood) from the corpses. Underneath the table, there is a sign that reads "Wear Protective Equipment," which is probably another remnant from the days when they used to treat contaminated corpses. It might be fun to have a drink at a table like this, though I'm not sure if the sterilization process is thorough in Night City [laugh].
Probably the gutter on the table (8:55-) is for the drains too; it was pretty depressing for us at the time because we had to look at all sorts of authentic photographic materials when we built this place [laugh].

gamesampo03.png
gamesampo04.png

Q. Where do you get such photos? (9;20-)
I usually search the Internet. I don't use them as they are of course, but search for images on Google as a way of collecting materials. I also use Pinterest a lot because it's very useful. If the game is set in the Middle Ages, like The Witcher Series, I would use books on medieval architecture, but in cyberpunk, there are many reference materials scattered all over the Internet.

Q. Where do you get your inspiration? (10:46-)
It varies. Not to mention the original 2020, but in some cases, each concept artist and 3D artist find materials on the internet and incorporate cool shapes in their work. And of course, a lot of devs like AKIRA and Ghost in the Shell, so they are often inspired by that.

gamesampo06.png

Q. Who decides on the lighting design and at what point in the development process? (11:35-)
To some extent, the direction of the design is decided at the stage of concept art. However, as development progresses, we all make changes little by little, so someone sneaks in from time to time and changes something, like "I think this is better!”
For example, the lighting in the Afterlife was done by a 3D artist, a colleague of mine, who was committed to it and made many changes to improve the quality. That's why some parts have changed from the original concept art. We also have a lighting team that specializes in placing the lights, and they make further changes to the placement of the lights every now and then. All these works add up to the final product of the Afterlife.

Q. What's the significance of the lighting in the Afterlife restroom? (12:59-)
Not sure. I guess… the lighting artist decided to make it this way? I don’t know [laugh].
In any case, there are four different styles of restrooms in Cyberpunk 2077. Of course, this is not limited to restrooms, but it's a whole concept that runs throughout 2077. That's why there are four different designs for smartphones, vending machines, and even toilet bowls. When it comes to restrooms, The Afterlife is Entropism style. The Konpeki Plaza, on the other hand, is Neomilitarism.

gamesampo07.png
gamesampo08.png

Q. A lot of stickers on the tables, walls, etc. are put up by each person as they please? (15:24-)
Well, yes. The 3D artists in charge just put them up.
By the way, the image data for these stickers and posters is very big. On the other hand, when it comes to objects such as concrete or steel plates, we only need to repeat images (called “tiling”) with a size of, e.g. 2m x 2m, so even if there is a 10m pillar, the texture image data is only for 2m. Stickers, on the other hand, have a different design for each one, so they eat up a lot of memory. That's why programmers get very angry when we put up a lot of stickers and posters. From our point of view, "I want to put up lots of different kinds of posters to make it more lively!” But sometimes we are forced to remove stickers to reduce the number of textures [laugh]. The Afterlife has quite a variety of stickers, but that's only because the programmers worked hard to make it possible.
CG is basically made up of "models" and "textures". A model is a 3D representation of a shape, which image data is not as big as texture data. Even textures won’t be a problem when they can be used repeatedly, such as wall tiles, but since Cyberpunk 2077 has a ton of textures with individually different designs like stickers, which causes so much data, it’s hard for us. From an art director's point of view, "This place needs to be filled with more graffiti!” But if the programmer says, "You're going to get in trouble if you do that”, then both parties have to find a happy medium.

gamesampo09.png

The tunnel to the Afterlife (22:19-)
This back alley leading to the entrance of the Afterlife was added towards the end of the development. Initially, we had to take a detour from V’s home to reach the Afterlife, but after receiving requests from level designers and quest designers that the Afterlife was visited so often that it had to be easily accessible, we decided to renovate the building and add the tunnel later.

・・・​

▶︎ Corpo Plaza (25:12-)


Q. Will there be any inconveniences when the teams in charge of each area create their own maps and combine them at the end? For example, roads and streets sometimes don’t fit together as planned? (25:52-)
It does happen sometimes. However, the Night City in this game is basically a collection of modules, and the buildings themselves are made up of many small parts as well. For example, if we assemble a building with a 6m wide by 4m high wall piece, and want to expand it later, we can do so easily by adding one more piece, just like LEGO blocks. The same goes for roads and sidewalks, which can be quite flexible in various forms.
Another way to create an open world game - which varies from project to project and company to company - is to model the building itself as a single unit (not as LEGO blocks as we do). For example, from what I've seen from the outside, Rockstar Games probably uses that method (GTA, RDR) . I guess that's the way that best fits their art direction and their idea of what kind of city they want to create. This time we needed to build a diverse and flexible Night City, so we decided to develop it in a modular way.

Q. From a game developer's point of view, what aspects interest you when you play other developers' games? (28:22-)
The first thing that catches my eye is the repetitive patterns. For instance, when I see the marble texture used in the staircase (28:45-), I wonder how this texture is used repeatedly. By understanding this pattern, I can estimate not only which parts are reused, but also the border and even the image size of the texture.
As I mentioned before with the stickers, image data is big, so repeated use of textures is an inevitable part of game development, but how often are they reused? Also, texture repetition is particularly noticeable in roadways. In those cases, sometimes another texture is added on top of it. We use a variety of techniques in this game of course, but how do other games hide it? There are many ways.

gamesampo10.png
gamesampo11.png

For example, the crack in the asphalt that you see now. That's called a decal, which is applied over the top like a sticker. On the polygon, it is completely flat, but using the latest technology (like parallax mapping), it gives you the illusion that it is really uneven.

Q. No matter how I look at it, it looks like it is concave...
In fact, it is not concave at all. It is completely flat. This decal is pasted on top of the asphalt texture that was originally there to hide the repetition of the texture.
Anyway, that’s what I can’t help but pay attention to the other game, How do other developers hide text repetition so that the player doesn't notice it? How do they try to make changes in the repetition? That kind of thing.

・・・​

▶︎ Designers VS Artists (33:54-)


In game development, this kind of exchange often takes place between the designers, who work hard every day to make the game play more interesting (such as a level designer and a quest designer), and the artists, who are mainly in charge of the game's appearance (such as an environment artist and a character artist):

For example, do you know why this traffic light pole is so thick? There are two reasons.

gamesampo12.png
gamesampo13.png
  1. Because when the player is driving, we need to make him/her sure that "this traffic light will not break”. In order for the player to be able to judge in an instant that it is unbreakable, it must not be halfway thick. That's why the designer must demand that the artist make the traffic pole thick enough to appeal to the player that it is unbreakable. Even if the artists think that such a thick signal is unnatural, they have to design it in a cool way within those constraints.
  2. Another reason is that the player must be able to instantly determine whether or not (s)he can take a cover position with that object.
    This criterion exists in all shooting games with cover actions, and they usually have a restriction such as “objects in which the player can take cover need to be at least one meter wide" (the number varies from game to game). If an object is slightly less than one meter wide, we can't make such a pole because the player won’t make sure if it’s available or not to take cover. On the other hand, if the pole is less than 40cm, it's fine because the player will understand immediately it won’t be the place. Even if the artist thinks that “this traffic light with 80cm wide pole would be the best”, it's very common that the pole has to be either 1m or 40cm. (in this case, there is no pole in the game that is less than 1m and more than 40cm)
    These restrictions exist not only in the width but also in the height, so the height of all objects that can be covered is set to "the perfect height” so that the aiming point is slightly above the object when the player aims. All these settings, which are designed not to interfere with the gameplay, are all thanks to the level designers.
It can be said that a large part of the work of an environment artist is to take into account the constraints of shape, height and width, and then make it look natural and realistic within those rules.
 
Last edited:
Episode 2:

・・・​

▶︎ Kabuki District (0:24-)

This area was requested by the level designer to make the entire rooftop available for gameplay. That’s why even though the area is complex with elevation differences, it is still a combat-ready space where you can jump across the rooftops and take cover without hindering walking. This district was difficult to create because we had the concept from the beginning of development to make it a space that would allow for three-dimensional gameplay. But in the end, I think we ended up with an interesting district.
Not only in the Kabuki area, but when creating a complex district like this, we start by calculating the scale of a mere cube, and then placing cubes of the desired size as buildings. Then, while considering the size of each building and the connection between buildings, adjust the size of the cubes and gradually decide on the layout. By the time the general layout is decided, the parts for each building will be available, and we can replace the cubes with those parts to make it look like a real building. At this point, small items such as cables and outdoor air conditioning units have not yet been placed. We actually play with the field of just walls and rooftops, and make further adjustments, refining the layout until it takes on its final form. Only after that is decided do we move on to the stage of placing small items such as cables.
This is how we build the city; since the necessary adjustments and verifications are made at the right time, even complex layouts can be done rather smoothly. The first step is to create a rough outline, and then work on the details.

gamesampo14.png

Long Distance Rendering (5:11-)
In a game, environmental elements that the player cannot see are not rendered. For example, if there is a large building in the foreground, another building hidden by it is not drawn. This is because the computer decides that the building is not currently visible and processes the data accordingly. Therefore, if the player climbs to a high place and can see many buildings, the game will have to display all the buildings, which will increase the load. Right now, your PS5 is just working hard on the processing. But since there won’t be a battle from the perspective of a distant view, it can be said that it manages to handle the processing.

Q. There is a cloud of steam coming out of the manhole. Does this kind of process cause being sluggish? (6:03-)
Basically, yes. Compared to asphalt, steel poles, and other things that are not completely transparent, translucent materials are slower to process because they require processing to determine how much of what is behind them is transparent. This is not my area of expertise, so I've heard this from programmers.

・・・​

▶︎ About Objects (9:13-)

In a game world, all chairs are of a uniform height. There is only one type of sitting animation (for that height), and the internal processing is such that you are not sitting on an actual chair, but like an imaginary chair. In Cyberpunk 2077, there are three different heights: "barstool”, "normal chair”, and "sofa”, all of which were specified by the animation team. That’s why the chairs we make have to be exactly the same height as these three types, with no more than a centimeter difference. Otherwise, your buttocks will sink into the chair, or you will sit in an unnatural floating position. The same applies to the height of the table, and there is a rule that when the chair is at this height, the table must be at this height.

gamesampo15.png

A cup placed on a table (11:20-)
The cup on the table is not actually placed on the table either, but the internal process is that the cup is stationary in the air at the height of the table. However, if it breaks or moves (due to the player's impact), the physics calculation starts at the moment of impact. In other words, calculating every time an object is stationary would only make the process slower, so the moment an impact is detected from a stationary state, the process switches to physics calculation.

This is a bug… (12:08-)
Despite shooting the table, a cup remains stationary in the air. The reason why this bug happens is because the cup does not know (internally) that the table has been broken by the gun. "Notifying the cup that the table below you (the cup) is broken" is a rather difficult process, and also requires linking the position of the table and cup one by one. In the case of this table, that setting may not have been done properly.

gamesampo16.png


・・・​

▶︎ Memory Budgeting (14:14-)

In game development, each team is fighting for the "limited processing power", which means detailed and proper processing can only be used in the most important situation. For example, we can't use the same processing power for mob characters as we do for combat. This is where our know-how comes into play, as we have to simplify (the process) as much as possible, while at the same time making it look natural. One such example is the fact that the mob character I mentioned earlier is not actually sitting on a chair, but the accumulation of such things is essential in game development.

Q. During development, do you have any information like "this is the amount of memory you have available“ as an actual number? (15:24-)
Yes, in the development screen, we can see various numbers as debugging information such as the current memory usage of each team. And on top of that, it also shows things like, "This is supposed to be their memory budget, but they are using extra!” in red [laugh]. So if you put too many stickers on the background, you'll see it in red and know right away that you're using too many textures in the specific area [laugh].

Q. As game consoles become more powerful, will the developers have the benefit of it? (16:19-)
It would be nice to have more memory available, but I'm not sure if it would solve the problem. For example, PS3 had 256MB of GPU memory. At the time of its release, all developers were saying, "How can we use 256MB? If we had that much, we could do anything!” However, the question soon changed to "How can we reduce the memory usage? In PS4, the combined GPU and CPU capacity was increased to 8GB, but just like in PS3, everyone ended up struggling to find the right amount of memory. In PS5, the capacity has been doubled from PS4, and so far it seems to be fine, but we don't know what will happen in the future.

gamesampo17.png

Q. You mean, even though game consoles are getting more powerful, there is more work to be done? (17:27-)
In fact, it is becoming more and more difficult, especially when it comes to the development of AAA games, where the number of developers is in the hundreds or thousands. At the same time, however, it is also true that today's procedural technology (technology for automatically generating complex data by setting rules for data generation) has made tremendous progress. In Cyberpunk 2077, there is a huge garbage dump in the suburbs, and I believe that area was created using procedural technology. The background artist created a model of a single item such as a garbage bag, and then simulated how it would spread if it were scattered in large quantities.
Houdini software is often used for procedural generation, and it is used in a wide variety of applications because it allows you to set the logic of what you want to simulate. For example, it can be used not only to randomly scatter something or automatically generate flames but also to create the building itself, depending on the project. The parts of the building have to be prepared by us, but in a case like “Build a random building from the second to the tenth floor. The number of windows is also up to you (Houdini)”, if you set up the logic of the rules, Houdini will generate the building accordingly, which is very convenient.
 
This game have a lot of environment (especially architectural) details that simply not being talk a lot more, because (of course) the current condition of the game & its receptions right now. Some unfinished buildings & empty rooms & empty mega buildings also have some interesting details. There were a lot of interesting stuffs in late game places such as Arasaka Tower, Hanako's mansion, Rogue's room, or even side stuffs like Hardcore Samurai's fan apartment room, some of the less known bars, etc, etc...

This is even unsurprisingly weird when the in-game map is actually different (missing so many stuffs) compare to the Night City promo/poster map or the Cyberpunk 2077 artbook map.
 
Q. You mean even though game consoles are getting more powerful, there is more work to be done? (17:27-)
In fact, it is becoming more and more difficult, especially when it comes to the development of AAA games, where the number of developers is in the hundreds or thousands. At the same time, however, it is also true that today's procedural technology (technology for automatically generating complex data by setting rules for data generation) has made tremendous progress. In Cyberpunk 2077, there is a huge garbage dump in the suburbs, and I believe that area was created using procedural technology. The background artist created a model of a single item such as a garbage bag, and then simulated how it would spread if it were scattered in large quantities.
That sounds neat. I don't really know about any of this stuff but I wonder if this will work with raytracing and if it's raytracing at some point, 5 to 7 years could help with production bottlenecks.
 
That sounds neat. I don't really know about any of this stuff but I wonder if this will work with raytracing and if it's raytracing at some point, 5 to 7 years could help with production bottlenecks.
Actually, I didn't know anything specific about this kind of stuff either, nor familiar with. But it was so much fun to hear an interview where the questions were answered completely from a developer's perspective. I hope everyone enjoyed it as much as I did. <3
 
Top Bottom