Creating a cyberpunk 2077 like city in blender Pt1 (Level of detail, concept)

 

I'm trying something very cool - to create a scene like cyberpunk 2077 in Blender. Here's my work in progress with comments so you can understand how I think.

As an architect, I thought about creating a high-rise building as in real life - with columns, beams, structures.. (I'm talking specifically about the structural part that sustains the building, that most people don't pay attention to) as you saw in these photos. 

Now take in mind taking this approach in an open world environment, and probably it would have a lot of buildings.. and let's multiply by the LOD (level of detail) and especially in this case Level of reality (columns, beams, structures). The result? 

This approach can consume a lot of memory, even though is correct. 

An example of where this approach is needed, also, is welcome due to the concept of the scenery/idea is below:

STEP 1: Columns, beams (structure)

STEP 2: Defining textures, direction, what will / won't repeat

STEP 3: Texturizing (using super texture plugin), defining Height / Displacement maps, etc.

STEP 4: Inserting walls, windows, openings, roofs, according to the structure. 

PS: I will create a post dealing with this work (medieval scenery) too,  created  with Blender.  I will write about memory / cpu and other things that must be took  in  mind when projecting a scenery. The full project is here: https://www.behance.net/gallery/109918271/Medieval-Scenery

Why this approach? 

The building in that time (at least I/We believe) were built that way, which the player  needs to feel like. Columns, beams, and some details are needed. I could neglect and make without that, but would seem flat/not too real, taking into account we are in new gen of consoles/pcs.

Ok, going direct to the point:  Here's my concept:

-Height: 250 meters 

-Width and length: 50 meters

-Steel structure Internal (not visible) and external (visible)

-Approximately 80 floors

-Helipad

Let's build it!

I'm using some basic colors to show the difference between objects.

Another idea is that the system in which this mesh will be used, could do the following things:

• Process details according to the distance from which the user is seeing the object, for example: from a far distance is not a good practice to draw the inside of a building.
  
  
• Maybe even use textures according to the necessity, distance... things that the human eye will see only when next.
  
  
• Use the same texture in different settings, only changing the colors or even quality.

To be continued...

Testing Armorpaint #1

 

As you can see, I'm testing armor paint here, especially on some objects I've created in my medieval scenery.

Here's the node setup for one of the barrel textures (rings):

The final result is not bad at all!  
(I'm also using the armor paint live link plugin). 

I'm also planning to test substance painter (trial) I will post here my thoughts about it. 

I can say, at least for basic things, you can do the same things in armor paint that you would in substance painter. For advanced stuff, you need to do SOME WORK (because the substance painter has some defaults parameters, so it makes fast).

Search for youtube videos showing what you can do with armorpaint. My tip is that you need to know how to work with blender's node system since AP uses the same idea.

A more complete screenshot: (I'm using super texture plugin again to save time)

Creating objects : Thinking in an optimized way

 

One thing I think is ideal is to create an object that is 'perfect' and, at the same time, does not overuse the target system's (PC, Cellphone, Console) resources. 

For example:

Whereas the 'bone' structure is really simple:

This object has:

Triangles:     4,334

Faces:             2,046

Vertices:         2,192

(more optimized looks like this)

Wrong way to do it: In my humble opinion: Too many lines and vertex: it's heavy and resource consuming (imagine the same object drawn a lot of times in the screen, not only once).   

Only because today's PC's and consoles are potent, does not mean that don't need optimization. Let the texture (bump, displace, etc) do the work when possible. 

Or, even, we can create an adaptable mesh, wich the system could depend on the situation, use the high-quality mesh and or low-quality mesh even reduce automatically, depending on the situation - such as:

I already said someplace - "from far distance there's no need to view the squirrel in one of the tree's branches, just a point moving can do the same effect, whereas as the closer you get the point will transform into a moving thing, then finally into a squirrel".

This is just an example. you must design/export meshes having in mind the export format that your enterprise/GameHouse tool(s) is using. Some prefer meshes with quads (quad geometry), others, with triangles (triangulate faces) when exporting. Avoid n-gons!