peter_spy

That quote is the example 2048 x 2048 tiling texture with default scale set to 1 fits on a wall that is 512 x 512. So, it's actually scaled down by 4. Every editor does it a bit differently. TDS editor is old and didn't scale textures at all. I think UT3 editor scaled textures by 2, UDK did the same thing. Of course this is for practical reasons, noone needs 2048 texture on a 2048 wall and texel ratio of 1 pixel to 1 editor unit Edit: 1) 2048 texture with scale 1 - fits on a 512 u wall: 2) The same texture with scale 0.5, it fills area of 256 x 256 units: 3) And the scale of 0.25, you can see the pattern because the texture area covers only 128 x 128 units: ...but it looks sharp as hell, you can basically stick your nose to the wall and it still looks great

8192 is it? I don't think I'll hit that, definitely not with this small project I was referring to TDS path system as something to use as inspiration, because it minimizes the clutter these path points make in the level (and at some point they always do) and it's easier to debug. Default patrol points we're linked to each other from start to end and used properties where possible (like for waiting, changing the walk direction or speed). Properties that needed to be in separate nodes because of the directional data (like LookAt points) were sort of "plugged into" certain patrol points (i.e. the link was from LookAt point to a patrol point). I wish I had TDS editor installed, screens would give you a better idea on that.

Alright, I'll keep that in mind, thank you. That will effectively double the number of path corners though. Not that I'm hitting any limits yet A thing to consider then: TDS had a property like TurnBack: [bool]. You put that in first and last patrol point and you had a nice loop. Fast and simple solution. FWIW, TDS had a decent AI pathnode design, clean and fast to use. A bit less "sequential", so any modifications were a bit easier to implement, especially with complex routes. How AI used it in real-level situations though, that's another story

Yup, I'm using wait points so guards take a break before patrolling, this works fine. But when they're patrolling, it looks like this: 1) This guard starts at the lower level, and he sticks to patrol points: 2) But this is how he walks back to his starting point: He hugs the wall to get there as fast as possible, he's not following the partol points.

Just a small update, the AI and patrol points are there, now the "gameplay lightning" phase. I got one question though: is it possible to make AI walk back a certain route and stick to path_corners? Linking the last node with the first makes the AI take the shortest route towards destination as it turns back, and it's not necessarily a good thing. On unrelated note, I knew the editor was lying to me about texture scale, as most editors do, but didn't know how much. It looks like the actual ratio with scale 1 is 0.25 (i.e. the 2048 texture fits on the 512 wall). This will be useful later, to maintain consistent texel ratio for modular meshes and unique objects vs tiled textures on brushes and patches. And, you know, the general sense of scale

This seems to be more about fields like music and music industry, I think. When it comes to games, editing and mappers, I think what Melan says is more accurate. It also confirms what Hourences wrote in 2008 in his book about level design: ^This has already happened.

This is kind of trying to water down the whole issue. The pivot/origin point has to be aligned with the mesh and the floor level, to allow basic flexibility and fast work, period. And it requires almost zero effort to do this. Move the pivot, go back to mesh level, move it back to 0,0,0 - done. Leaving it the way it is now is simply a mistake, it undermines the concept of modularity, as mesh alignment is harder and takes longer than necessary. This really is the basics of that concept, and arguing over this is pointless. It's like wondering whether water really boils at 100 °C. And trust me, mappers who will come here from other communities are very familiar with the concept. Games from past 15 years shipped with editor have been using modular stock meshes. Again, this is not personal attack on Springheel, I'm sure he'll do better next time. Besides, there's nothing wrong with making mistakes, how else you're supposed to learn? Pretending that you're not making mistakes is worse IMO. Btw. right now I'm trying to design modular meshes for my first location, and I know how hard it is to do it properly. I already spent a lot of time taking screenshots from e.g. Dishonored and browsing through UT3 maps to remind myself how to design possibly most flexible tileset, that will work in all scenarios I can think of. Sorry, AluminumHaste, I promise this is my last post in this topic. I just thought it was worth to clarify that. If you find this offensive (and I really don't mean it), feel free to delete it. I'll be making my point in more constructive way from now on, showing you my modular meshes in the WIP thread.

I think you're a tad too sensitive, Gray (edit: sorry, somehow I thought it was Biker) Sure, players won't notice that, but we were talking about mappers. A good and fast system of pivot/origin points shaves off hours and days of mappers' work. But I agree with AH, there's no point beating a dead horse, what's done is done. There's always next time.

Hey, you don't have to sell me on the concept of modular meshes, I've been using those in several Unreal engines throughout many years That document I linked to, it's from 2002, and it's been around the web probably since UT 2004. This isn't arcane knowledge. People who are or will be drawn to TDM and DR after Steam Greenlight success will come with different backgrounds, like Unreal or Unity. They will see your models and they will wonder whether you knew what you were doing. It would take like a second per model to fix those pivots, and yet they are in weird places right now. I think it's kind of obvious that there's a slight difference between meshes made for your own purposes only, and those released to the public. The former can have flaws unless you know how to use them and you're satisfied with the result. For the latter you need to come up with a system and be as consistent as possible so others can recognize the idea behind this. Yup, that's what always happens with those. They will move in relation to the pivot point on the map.

Heartbroken? I think that's way too dramatic Me ≠ my work. I don't feel like that time was wasted, quite the contrary. Me = my job, that's probably more like it

I fully subscribe to what you just said. That is the main reason I never finished three TDS FMs I had in progress, at least one GoW map, and two UT3 maps But I still value that experience, It helped me grow, even if noone really saw it. I know your job, home, and other adult responsibilities leave little time for editing. But, things like properly placed pivots are kind of minimum requirement to meet, if you want the modular system to work. There are free resources on the web to help you, like this one: https://docs.unrealengine.com/udk/Three/rsrc/Three/ModularLevelDesign/ModularLevelDesign.pdf Also viewing and analyzing meshes in other game editors may help as well, like UT3 is pretty cheap now, it comes with the editor and mesh libraries. Anyway, despite the language barrier or tone problems, my attitude is always "here, let me help you improve this", not "your work is shit".

That kind of depends on your goals, and the so-called peer culture. If you value self-imprevement most, you will receive even the most harsh critique. You might be angry at first, but you know it's valid, even if poorly-phrased. Some communities (like Thief community in general sense) are very polite and encouraging, there's barely any critique. On Polygon forums, people are usually polite, but they will point out your errors, one by one. Last but not least, there are forums like max3d.pl, where people are merciless They will tell you your wireframe is shit when they see it, and they will find every unnecessary polygon you may have missed, every wrong seam, and every distorted UV. And that is good too, as long as you realise they point their critique at your work, not at you personally. You always release your work for at least three audiences: the general public, which is typically the most generous in their appraisals; your "specialist colleagues", who will analyse your work at every angle, and try to learn your technique, if it's good (or tear it to pieces, if it's shoddy); and finally yourself, to fulfill your creative needs or find answers or solve problems nobody else even thinks of. The first audience is nice to have, but the other two are true source of your self-improvement and development as a craftsman, artist, whoever you think you are.

I didn't mean to offend you, the whole initiative is a great idea. But, I'd send those meshes for a peer review and listen to feedback before publishing. E.g. my work is rarely a result of a first try. Stuff I'm really happy with is usually the second or third version, often made from scratch after hitting a dead end and receiving a valid critique from my architect friend (who also plays a lot of games). I mean, players probably won't notice that. Fellow mappers and modelers, they will be much, much more harsh in their judgement. I haven't tried blender, but 3dsmax won't import lightwave files. I tried to convert it to ase or export to obj, but apparently this script works only with brushwork.

I started fiddling with building pieces and mansion parts, and the first impression is... kind of messy. Pivots for walls seem to be in a wrong place, they should be at one of the wall edges, or even better, at a corner, to facilitate rotating. Mansion walls are impossible to snap to grid. When I align them with the ground brush, the pivot is somewhere in the air. Such big architectural parts should be easily manipulated within grid of 16 or 8, in all directions. The other problem is the meshwork, there seem to be unnecessary polygons and crossings here. Frames on the other hand, are long and have too few polygons. Lightning on long faces with only two triangles will look ugly, especially with specular. And, assuming that I'm using those meshes correctly, they're not flush with each other as well. Column base trim texture seems to be inverted and the geometry of columns is off at every wall section. Also, quite a few of these walls have faces on the other side that won't be visible. None of these walls is thicker than 8 or 12, so I assume wall thickness was left to mappers decision, as it should be.

I haven't tried slopes steeper than 45°, but both player and AI seem to handle that. AI looks a bit weird though, the way they bend their knees looks kind of creepy. I'm trying not to go over 22,5°, which is basically a slope for 8x16 steps. I haven't tried any spiral staircases yet, I guess those might be a problem with slopes.

Yup, that's why I used this phrase. TDM accommodates a bit for steps that are 8u high and 16u long for player movement, mostly when going up the stairs. Going down still looks jarring, and the AI looks awful traveling in this zig-zag motion, whatever the direction. Sure, the slope is a tradeoff, and sometimes things look a bit floaty, but it's still less distracting than blocky stairs.

Speaking of walkways, whether using mesh or brush for stairs, as a rule of thumb they should have a sloped collision model to avoid that jarring screen shake when player goes up or down. Typically our brain ignores that kind of movement when we climb the stairs, but not on screen

These are gameplay-oriented tricks, not standard practice. Doom 3 might have used it in some sections, but neither Thief series, nor any other non-stealth games use that on a regular basis.

Well, that's what the color range standard is for. It will look a bit washed out on every screen, but there won't be any pitch-black parts, and users can adjust brightness and contrast as needed. All games and movies use that. Some games use that night vision system too (like Thief reboot), not instead of ambient light setting, but as a complementary measure.

That ambient lightning part might be a bit confusing for newcomers. Ambient lightning is good, not only because it brightens the pitch-black areas. It also works like color range standard for movies and television (e.g. movies never use 0-255 color range). It allows similar experience for people using different brands and models of hardware. There are millions of combinations of monitor brand, model, Windows color profile, graphic card driver and settings, and the in-game gamma setting. Ambient lightning is the easiest way to take it all into account.

Btw. long ago, there was a mod for Mount and Blade called Polished Buildings, the author made an excellent tutorial on modular buildings. English translation is a bit rubbish, but you'll get the idea: http://wiki.polycount.com/w/images/0/09/Modular_MountBladeMod_01.jpg http://wiki.polycount.com/w/images/a/af/Modular_MountBladeMod_02.jpg http://wiki.polycount.com/w/images/9/93/Modular_MountBladeMod_03.jpg

That's 6 materials. I guess apart from the two decals which probably use one texture per material, the rest has to use at least two 2 textures per material, glass probably uses more (cubemap etc). You can be efficient with big textures, e.g. you can divide your mesh into interchangeable parts, like in the exploded view. Of course that requires a bit more planning. E.g. these parts use one material per set:

Well, that's the usual workflow, you have to maximize the efficiency of 0-1 space and yet allow for some flexibility, but with as few textures as possible. On that note, I wish unwrapping wasn't such a chore, it almost feels like punishment for the joy of modeling...

What I meant is this window uses 6 materials, so at least 6 different textures (probably more like 12, e.g. diff+normal, diff+spec): While in other engines it's more efficient to go like this: One mesh, one material, diffuse, normal, spec textures. Material instance for lit version with additional mask texture. As you see, this one also uses mirrored UVs to save space, which is a very good idea (not sure whether older engines support that). AFAIR, using multiple textures and materials increases drawcalls, impacts loading time, and wastes video card memory, especially with non-square textures. To minimize the impact, I always used one texture space for the whole model, and one material, if possible. Two materials (glass, stone) for the window would be fine, 6 seems a bit too much.

Question about the multiple textures/materials on meshes: I was taught that using multiple diffuse and other textures for one mesh is generally a bad idea, performance-wise. Meshes from UE-based games like UT3 or Gears of War use one texture space for e.g. window frame + glass. Things get a bit more complicated with Thief/TDM and multiple surface types, but I always addressed that by using the same textures in cloned materials, just changing the surface type and assigning the material to proper parts of given mesh (wood, glass, stone etc). In TDM's mesh browser however, I see simple meshes like windows having multiple tiling textures with 5-6 materials, with stuff like dirt or AO as a separate texture and material. Are you sure this is efficient?