Sotha's mapping thread

[Tels]

So what just happened is I built an *entire* ground floor of a castle using the premade modules. It took one hour. Most of the time was consumed by manual visportalling as, in my haste, I forgot to include the visportals into the modules themselves. After that, I added the VPs to the modules of course.

 

Here is my favourite spot:

http://i.imgur.com/APgwR.jpg

 

Now the plan is to build the full castle geometry using the modules. Later I'll add detail modules I can overlap onto the existing modules. A detailed ceiling light, for example. Once it is ready, I'll clone it into the geometry modules I need it in. I'll try to do the same with dirt and other aging effects. Lets see how this goes. At present I'm very confident in my progress.

 

And after a while you will discover that "redoing the modules is a pain" (after you build your entire castle and noticed you forgot some minor detail or made some minor mistake) and then you probably will see why my "build prefabs, assemble them automatic, refine them then assemble again automatic" is easier in the long run then your way Come to the dark side, we have cookies

[Sotha]

What kind of cookies...?

 

Anyways: I've never been into rebuilding as a mapper, so if I'm displeased in my modules, I'll fix the accumulating map by hand and fix the modules for the future. After initial experiments, I'm now settled with the modules and they seem to be perfect for my building needs. All it basically needs is learning the requirements and the correct work flow.

 

Building the modules I was thinking about the automatic placement and there are some issues. Texture alignment for example two neighbouring module floors. That is difficult to do without getting ugly seams with some textures. It is easy to fix by hand, just choose every floor module and set natural texture scale. Also by hand I can add charming little details at will: move this thingy a bit over here, and this other thingy over there. With autoplacement you need always a new module for it.

 

But anyhow: once I get these done and release them for everyone, your autogenerator thingy is probably in the state that you can start your own little map factory after you tune my pieces so your program can handle them.

[Tels]

What kind of cookies...?

 

Dark chocolate cookies

 

Anyways: I've never been into rebuilding as a mapper, so if I'm displeased in my modules, I'll fix the accumulating map by hand and fix the modules for the future. After initial experiments, I'm now settled with the modules and they seem to be perfect for my building needs. All it basically needs is learning the requirements and the correct work flow.

 

Building the modules I was thinking about the automatic placement and there are some issues. Texture alignment for example two neighbouring module floors. That is difficult to do without getting ugly seams with some textures. It is easy to fix by hand, just choose every floor module and set natural texture scale. Also by hand I can add charming little details at will: move this thingy a bit over here, and this other thingy over there. With autoplacement you need always a new module for it.

 

Not nec. a new module. First, one can combine many different modules on one place. For instance, here are the modules that get combined in some of my spots:

 

* empty (ceiling, four corner columns, floor), player_start_east, wall_east, wall_north, wall_west, crate

 

Now this results in an U-shaped room with a crate and the player.

 

OR:

 

* empty (ceiling, four corner columns, floor), pit, wall_north, wall_south, slider_east

 

This results in a pit with a torch on it and a slider covering it.

 

if you want to vary this, you can remove the crate or torch etc, replace it with something else, and also youc an vary the walls. Since the walls are specified extra, you can combine "plain wall", or "wall with painting" or "wall with wall-hole", or "grate in wall" or whatever.

 

Basically, this allows you an almost infinite combiantion of pieces with only a handful of prefabs.

 

And that doesn't even cover the (planned) randomization, where either the prefabs itself are randomized (e.g. the type of wall that appears is choose from a list of walls you specify), or you have a list of entities that are selected.

 

The "texture alignment" is an issue (my generator currently screws up the texture alignment, which doesn't make it easier, either

 

Also, there are problems like "every module has to fit almost every other module" and this kind become quite hard. OTOH, with some basic building blocks and some "random decals and enhancements" you can already add a lot of variations automatically. And some sets of prefabs don't have to fit other sets (the "dungeon set" and the "sewer set" can be sep. bundles).

 

But anyhow: once I get these done and release them for everyone, your autogenerator thingy is probably in the state that you can start your own little map factory after you tune my pieces so your program can handle them.

 

In my case I settled on 192x192x192 pieces and in my case the "base" (floor, ceiling, corners) is sep. from the wall. It might be possible to rework your prefabs, but I have no idea how much work that is.

 

Currently I am only building prefabs to test the technology, the code has priority. (Basically, "make it work before make it pretty

 

 

But yeah, I see so many similiarities, it is just that I think the entire "pretty it up in DR" is a dead-end because once you have touched the level with DR, you can't go back. (Excuse me if I repeat myself here, please ignore this :)E

 

Edit: I make some screenshots tomorrow to show the differences, because I think this is interesting to compare and more easily to see visually.

[Sotha]

Tonight I had only 45 mins to play around with DR. In this time, I built the second floor of the castle. I already had a rough plan what it should be like. Then I simply took a measure what kind of modules need to be combined, combined them all in the side of the map (built the entire second floor from the modules, that is). Then I simply drag-selected the second level and moved it into place. Got one leak because I didn't realize nodraw solids don't seal against the void. Also got a single wall of z-fighting, because of an error in door-module placement. Overall, the workflow runs rather error free now. It is actually more easy to do mistakes with the conventional mapping style.

 

I can already see that this is going to be the best-looking map I've ever produced. This is also starting to be more than just an experiment. A plot has started to crystallize in my head... If it works, I'm gonna do something different with this mission.

[RPGista]

Sounds great, experiments with storytelling are exactly what we (players) need and there aren't enough castles missions around (can think of only one, House of Theo).

[SeriousToni]

I'm in for the random house generator!

[Sotha]

A little progress report.

 

Had a pause in mapping for the duration of DXHR plus missing link. Recently continued mapping again. I've been building more areas to the castle, did some AI texturing work and Tels & demagogue helped me to overcome some last technical problems preventing me of achieving my intended goal. Now there are no technical obstacles anymore and there are nothing which I am not sure how to accomplish it.

 

The basic level geometry is mostly done and a few rooms have been detailed. I spent some time to hone the skills I learned from the road building tutorial and did something organic to the mission.

 

As for the modular build principle, I still obey the basic grid rules and it saves a lot of time as corridors can be quickly built. I also built a few test grime-modules. They are optional func_static decal modules that can be placed onto a similar module thus griming it up with a simple copy-paste operation.

 

It seems that the grid&modular way works well for interiors. I tried a few times to build some exterior areas, but sadly the modular principle fails there: I simply cannot figure out how to stick to the grid and how to lay things out. Because of this, the exterior was built in the old-fashioned way, but I left a modular reservation for a few buildings.

 

This mission will be the most plot-intensive mission I've ever made. I got the plot in my mind as fragments and I need to figure out how to carefully lay it out in front of the player. I would like to give the player critical no-brainer options that affect the mission outcome, but getting that done in an intelligent way is a challenge that seems to get more difficult the more you spend time figuring it out. So no guarantees in that field, yet.

[DeTeEff]

I also built a few test grime-modules. They are optional func_static decal modules that can be placed onto a similar module thus griming it up with a simple copy-paste operation.

 

This sounds interresting. Would like an further explanation on how you do it...

[nbohr1more]

Glad to hear you're expanding some new technical horizons Sotha.

 

I will be watching this thread like a hawk

[Sotha]

 

 

This sounds interresting. Would like an further explanation on how you do it...

 

It isn't very complicated. Imagine a 144x144 corridor module. It has:

two worldspawn walls

Worldspawn ceiling

worldspawn floor.

Func_static details (a vaulting or a window niche for example)

 

The square module can be copy pasted into a chain to form a long corridor. You need separate dead end module to close the corridor.

 

Everything is nice and tidy and now you want to make some grime. You examine the clean module and put some dirt decal patches onto the floor, walls, the detail func_static features and to the ceiling. You apply some general grime onto it and make all the grime decals into a single func_static. Now you can clone this grime thing and apply it to all compatible modules. Once in place, you can adjust or change the grime texturing by hand on individual modules so that exactly the same grime pattern is not repeated on nearby modules.

 

Yeah, it is still handwork, but it saves a lot of time since you get at least all the module decal patches in place with just pressing space and aligning the decal func_statc in place.

[Bikerdude]

that only issue you'll have with those FS Sotha is the vert-point drift each time one is moved. The mapper will have to revert to WS, snap to grid etc.

[Sotha]

that only issue you'll have with those FS Sotha is the vert-point drift each time one is moved. The mapper will have to revert to WS, snap to grid etc.

 

I don't understand what you are talking about? 'vert-point drift?'

You keep grid 8 and move the func_static decoration to a next module. Works: no WS revertions and no grid snappings.

[DeTeEff]

Ahh, then I get it!

[Sotha]

Tonight's excecise was to figure out a handy way of using the modular approach for exteriors. Here's the conclusion:

1) Stick to the 144x144 grid.

2) build an ugly caulk house (first figure, left side) and pretty high-detail exterior modules.

3) snap the exterior modules onto the caulk house surface (first figure, right side).

 

 

The end result looks like this. There are two different modules: the log house wall with a window and the door entrance seen on the side.

 

The problem, I think, is the amount of tris those logs generate. I probably need to reduce the patch tesselation even further with the path inspector to ensure smooth operation on lower end computers as well. If I get bored of the log house 'skin' the house has, I could build new facade modules and splat them on the exterior without redoing any worldspawn geometry work! Fun, is it not?

 

I'll report later back to tell how the interior modules went to place. Since I am on the grid, theoretically, I think I'll get away with the interior by simply cloning the exterior modules, turning them around and placing them inside.

 

Oh, and if one would bother to make the decoration modules into an .ase model, the map would use less memory as each decoration model is loaded only once!

[ungoliant]

since models will be func_static by default, that should fix the tri splitting of the logs, if memory serves, since nothing will be optimized out of it. just converting to func_static should achieve the same result.

[RPGista]

I like how the house experiment is coming along, cant really comment on the internal organization as it isnt shown but I like the fact that there's enough detail and the wood walls are actual geometry, not a flat plane with heavy bump maps. Most of the times, we are talking about two elevations per accessible building (front and back, if you think of tight urban areas in maps like Saint Lucia or Return to the city), so I would probably say a basic "lexicon" of interior prefabs, if I understand your intention here, could definitely be the most important thing (and potentially very useful) for most users.

[Sotha]

Yep, all the works are available to everyone when I release this map.

 

The most important aim with these modular blog entries, is to give other mappers some pointers: you do not need to build everything by hand if you just modularize a bit and use some time to think how you set things up.

 

If every piece you build is reusable, the mapping process should be greatly expedited: more missions, easier to build.

 

This is used in commercial games too: just look at DXHR toilets. Exactly the same toilet model room used everywhere with or without accessible air vents. All the mapper probably needed to do was to smack the toilet into place and connect it to the corridors.

 

So this is not about giving the building blocks to people. It is more like showing people that if they change their workflow a bit, modularize and build their own blocks, their mapping projects will be easier and faster.

[Bikerdude]

@Sotha

 

In addition why not make the whole building a model and put caulk inside the walls to seal it.

[Obsttorte]

if you stay with the module mappers could use them for different building layouts

 

one/two/three floors, L-/T-shaped base and so one

[Sotha]

Does someone remember the name of the spawnarg that controls the AI walking speed? I need to slow an AI a bit. Thanks!

[Obsttorte]

the wiki is your friend:

 

anim_rate_X, where X is the animation to play

 

x=walk for walking speed, X=run for running speed

 

http://wiki.thedarkmod.com/index.php?title=Speed,_AI

 

I didn't tested it so far

[Sotha]

I'm experimenting all kinds of weird stuff. Is there a way to make AI completely non-solid? Player walks through, moveables fly through them and so forth?

I've already tested

"solid" "0"

"cylinder" "10" -> "cylinder" "1"

and

"size" "32 32 68" -> "size" "4 4 68"

but none of them achieve what I'm trying to do.

[Obsttorte]

In "A night to remember" the ghost walks through walls and so

but that's maybe scripted

Edited December 16, 2012 by Obsttorte

[Sotha]

In "A night to remember" the ghost walks through walls and so

but that's maybe scripted

 

I thinks that is ordinary AI walking through a fake wall... I'm trying to make AI non-bumpy and non-solid for the player and moveables.

[Baddcog]

@Sotha

 

In addition why not make the whole building a model and put caulk inside the walls to seal it.

 

Main problem is that everything will render all the time.

 

And if the center of the model is inside the house then the entire outside will render when inside.