Merricksdad's Black Hills Tileset (First Look)

[Legacy_MerricksDad]

               

I think I have come to settle on a rock face which is less bumpy. Any additional bumpy can be added with placeable rocks specific to the tileset. Unfortunately, my less bumpy tiles are having some issues with the seams. I'm trying really hard to make the tilesets seamless, and not obviously repeating. I hate that part about NWN. Everything, even rounded stuff, seems so squared off. That is a major failing of tileset based games.

 

I'd like to work around that issues by making a lot of groups. In the previous iteration of this tileset, I had a lot of multi tile ramps. Some were long and slowly climbing. Others were short and used a switchback method. Others used cut rock face stairways, being able to maximize realistic climbing of a 45 degree angle. The more groups, the less tile like the set can become.

 

If only I had saved the mod data for the first image, I would like to use that again and modify it slightly. Maybe I can look back in my numbered saves. I usually save just before exporting, but I can't seem to find it at the moment. I'm crashing GMAX a lot playing with large, shared modifier stacks.

 

I've also improved the texture one more time. Adding more graininess to the previous texture, and darkening it back down a bit, makes it play much better with the vanilla area environment settings. It also looks better with the grass color.

 

Sorry I didn't release pics today as planned. That search for a computer, plus some paperwork, as kept me overly busy. We'll just plan to move fast forward with this on Tuesday...my original plan.

               
               

               
            

[Legacy__six]

               

While people seem to want to push NWN to be more and more naturalistic and realistic, I find that the less realistic stuff is the more charming it is (to use a six-ism).

six-ism? A wha?

 

 

When I first saw this thread I was like dayum girl this set is whack. But having seen this latest screenshot I reckon you're onto something really good. Angular is nice, but I think you get more visual mileage from having these big surfaces with long, pronounced angular edges, than you would from having lots of smaller jaggedy bits.

 

Texture is way nicer with the more natural colouring too. Though it seems a bit lacking in range - it has that over-edited-in-Photoshop look, where all the colours look like they've been compressed IMO. I used to find that rather than play around with textures too much, I'd often be better just going back to photosource and making only minor edits to get rid of the uncharacteristically realistic look.

 

*wildly opines without thought for the consequences*

               
               

               
            

[Legacy_MerricksDad]

               

Here's a quick screenshot of the updated texture, and smaller texture tiling on the mountains. Other than the green moss (which in the black hills is usually blue lichen), I don't think I could get much closer in game to what I was actually going for. This combination of real cracks and texture-implied cracks seems to be the most real I can do without setting the face count to "ludicrous".

 

 

Opposite what I suggested yesterday about the lower poly, less pushed out rock faces (which I liked at the time), going back to my four images that I posted caused me to want that shape back again.

 

To counter the edge issues, I have determined that the best way to cover the seams, which are really just mistakes my quick method is making during the noise and push modifiers, is hand modifying (and maybe later via script) the right side of any tile. By right, I mean any time a terrain leaves the right side of a tile. So if you have a tile with level 0 on the top two corners, and level 1 on the bottom two corners, what I'll do is modify that right side so that it hangs into the tile to the right. In the case that more than one side leaves the tile to the right, then the right and top will be modified. This allows for complete hiding of all seams, and also randomizes the non-walk space on all mountain tiles, so you can help avoid that repeating tile look.

 

The base tiles I am using have three varieties of each, but additional increase of those base tiles to 6 total might also help with more randomization of neighboring tiles. Can you tell I hate seams?

               
               

               
            

[Legacy_s e n]

               

if you're referring to seams caused by engine lights in the toolset and in game I suggest you to use the "lord Rosenkrantz" method, you can easily see by importing a couple of tiles of his rocky mountains tileset in max. basically the method consists of using a frame of verts inside the outer tile vert contour, where the z and the parallel axis values of each frame vert is the same of the corresponding edge vert (1 frame vert for each edge vert), the other axis frame vert value you can safely move to give variation to the tile, but the others should be anchored. this is so far the best method, even it still causes seam issues with engine lights, when the tilt between verts is within certain values. we makers had discussions about that behavior and found no solution. but for small tilts, or for tilts near 90 degrees the engine looks fine. that said, the effect is more visible with whiter lighings and textures, and you can disguise it with dedicated rims

               
               

               
            

[Legacy_MerricksDad]

               

That was the old method I individually came up with too, however, that method has one fatal flaw for me: I am constrained to the 10m square, so anything I want to have carry from one tile to another still needs to fit within that tile. Unfortunately, that method is exactly why I am trying new ones, because it does not fit my need to make the tileset less tile like. That method is still what I will be using for the walkable parts, or other flat tile-to-tile regions.

 

Very specifically, I want a tileset where the edge of one tile is not required to have the same edge characteristics of the adjacent tiles. I've been able to pull that off, and easily, with the 2m raised grass with rocky face. That rocky face is extended into the region occupied by the adjacent tile, and then bent in such a way as to subduct into the rock of that adjacent tile. Because the quantity of variant tiles, this multiplies the possible appearances between any two adjacent tiles. It simply requires that I don't fully detail, or make walkable, the outer 1/16 of any non-walkable tile edge which continues into the next. I then lose about a meter per tile, but that is not the most valuable real estate on the tile.

 

I find that the only problem with this method is that, IF I ever intended to do an exact walkmesh which makes clicking a non-walk area precisely defined (mouse pointer location), I cannot supply the precise data for that non-walk region within 1m of the tile edge. But again, I don't want to do that, so it matters not at all. I generally optimize and de-tesselate all the non-walk areas anyway. AABB calculatoins can be needlessly large when not compressed. I recently pulled off a 3.6 meg WOK file.

 

As you've mentioned, the only areas which will have obvious lighting flaws, will be the areas where I actually do use the same-edge method, and then only when lighting is too bright to let the engine cover it naturally.

               
               

               
            

[Legacy_MerricksDad]

               

Here's an example of what I'm doing. Obviously, there are two tiles per example, with the first example having an orange section where the seam may not line up well normally. Within a percent of the tile in either right or left from the orange line, the tile must sacrifice a region so that it can match up exactly with the adjacent tile. It must match both position and angle of change. But not only at the point of connection. It must also carry that angle of change a distance, and over at least one face long enough to make the overall smoothing curve match from tile to tile. Two small polygons to both sides of the orange line would improve the smoothing curve match. Three would further improve it, and so on. The shape of the seam will be exactly the same no matter what adjacent tile is placed.

 

 

In the second image, the right-exiting portion of the non-walk section carries into the next tile. The distance is similarly important to the previous example, except that there is gain to variance, instead of loss of variance. The distance carried must be enough to make the overall connection look realistic, but not so long that it interferes with the usefulness of the covered portion of the adjacent tile. Carrying 50% across the tile may visually block a walkable section, such as a level-to-level ramp. The maker must then be sure that any such ramp, or other walkable/interaction-intensive section of a tile is not placed within that intended shared region.

 

The green line shows the placement of the new seam. Instead of the potential of having a vertical line from the top of the crossing surface to the bottom, you instead have a zig-zag line which will "randomize" depending on which two tiles are placed adjacent. The red line shows where the edge of the subducting chunk will appear, or actually not appear, since it will be hidden behind other drawn faces from the adjacent tile.

 

The third image shows what it might be like to have a walkable crosser, and a non-walk corner terrain.

 

You might also take into consideration wrapping the corners of non-walk corner terrain to 270 degrees, rather than 180. This will cover two rights, the tile right, and the tile top. With the example above, wrapping both rights in this manner would require two 270 degree modifications.

               
               

               
            

[Legacy_s e n]

               

well I'm no big fan of that method to extrude the mesh beyond the tile edges, I think the same Lord Rosenkrantz and also Zwerkules used something like that for rocky terrain of variable heights, while it is true that adds variation and disguises the seam (no way to get smoothing there, and thats what I was referring with the previous post, for grass and also for smooth rock) it is also true that with 3 or 4 tile variations and a few placeables the builder has the tools to make an area "unique".  the real limitation here is given by the engine itself, especially the way it handles the multiple raise values, thats also the reason why I stopped working on terria (along with the emitter pierce through fog issue): there is no way to get the tileedges working for certain values, and the other method like in bloodmonkey's rocky mountains is deadly boring because it uses 1 terrain each raise value, so for 3 raise values you need to work with 4 different terrains just to get the rocky terrain, not saying there is the need to set a high value for the raise, and it effects all other terrains which would go fine with 2 or 3 meters raise value instead of say 15 or 20 meters...

anyway i see the multiple raise the only way to get the look of a "true" mountain terrain in nwn, with steep raise that goes in hand with gentle hills

               
               

               
            

[Legacy_MerricksDad]

               

well I'm no big fan of that method to extrude the mesh beyond the tile edges

 

But why? I find it works very well with certain texture and shape, especially the black hills granite with the multiple vertical folds. The method also works with out-of-tile trees. I commonly use walkable trees. If the tree is under so many inches thick, in comparison to the size of the player model, I allow passing through that tree. just like dungeon siege.

 

anyway i see the multiple raise the only way to get the look of a "true" mountain terrain in nwn, with steep raise that goes in hand with gentle hills

 

You'll see this tileset uses a few different methods to portray height. It has the mountain shown (10m) be the "raise" value. It gives two crosser types to be combined with two adjacent raise uses, making a sheer 20m cliff. It has two 5m terrain types which fit in between the "raise" and previous level. And it has two "slightly raise" terrain types, for 1m and 2m respectively. At this point 2m can be banked up against "raise", but not 1m. Right now the 1m cannot be banked against anything. When finished, all the raise types will be placeable near the higher types. So you'll have various contoured slopes leading up to jagged mountains: 1m, 2m, 5m, 10m, 20m. I may go so far as including 11m, 12m, 15m, 21m, 22m, and 25m, so that when a user places 1, 2, or 5m terrain, it won't stop them from attempting to use the raise feature. Instead, it will be simply fluid and modify as needed.

 

There is another method, one which I have used before, and very much liked. Instead of letting the tileset handle the "raise" feature, you use a GFF editor to change the height of tiles in the area file. This method is not interchangeable with the toolset tile placer, but once you do use the GFF method, you can still open it in the toolset and place objects, or even modify tiles as long as those tiles are not touching something the toolset cannot understand. Ive used a total of 8 height types and then modified their placement with the GFF editor. So I can have 3 height transitions on a single tile, just like dungeon siege, AND I can incrementally raise beyond one level. I made all increments multiples of 2m, so everything lined up really nice.  That uses a single terrain for all levels, so the groups you can place on level 0 work on all 8 levels.  To do that in the toolset, you'd need A LOT of repetitious tiles, and that takes up a lot of space and load time (same group for every level used). The GFF method saved me more than 75% of the model work I would have had to do. Somebody should alter the toolset so that we can use multiple increments at once (from 0 to 2 or higher), and then update the set editor to follow. The engine works with it just fine.

               
               

               
            

[Legacy_rjshae]

               

Could a set of rock placeables be used to conceal the seams?

               
               

               
            

[Legacy_MerricksDad]

               

In many cases yes. They would serve the same purpose as the wrapping, and probably look the same. I've entertained that idea as well, but since it would be required on every single tile seam, you'd need some variety, and the builder would have to place them all. So I said to myself, why not just put them on the tile?

               
               

               
            

[Legacy_KlatchainCoffee]

               

Just want to say - MD, you're amazing. Each time I think I may finally part ways from NWN for whatever reason, folks like you stop me by creating some new awesomeness.

 

Thank you! '>

               
               

               
            

[Legacy_MerricksDad]

               

ok, so here is this morning's proof of concept shot. Below you can see that the hobgoblin is standing at the line between two tiles. You can also see that, even though the tiles have shared heights at the walkable areas, they do not share the same height at the extruded area meant to be non-walkable. The image on the right shows the true boundary between tiles in orange, and shows in green the visible face boundary of that tile, in relation to the other tiles' visible faces.

 

 

This is done with a similar method to how children make clothing for paper dolls. Below, the image shows how I added two 1m face columns to the edge of the tile.

 

 

With those extra faces selected, I apply the list of modifiers (listed on a previous post in this thread). The result is what you see in the left of image above.

 

Bright light highlights the seam, but since other seams exist on the same rock faces, which are not the tile seam, it looks natural.

 

 

Edit:

 

While applying these changes to the dilate and erode crosser regions, I find that when I need to make a paper wing, I must also include down. This is because dilate and erode crossers allow a tile to carry directly into the tile above or below, making a 20m cliff. So to fully "paper-doll" a tile, you need right, top, and down. As shown in the image above, you also need any mirrors of those coordinates if you have diagonal tiles.

 

With the dilate and erode crossers, when I apply a wing section to the bottom, the wing is only one poly long, but much longer than the 100cm poly I used in the image above. This gives a very long crack from top to bottom of the tile when noise is applied. For weathered granite, this looks very natural.

               
               

               
            

[Legacy_3RavensMore]

               

Irregardless of the arcane magic that you use to create this, it looks to me as the most realistic example of mountains in NW to date.  As one who loves the Black Hills, and mountains in general, I really look forward to building with this stupendous tileset. 

               
               

               
            

[Legacy_s e n]

               

I like the texture although I would cut some purple and leave space to more gray, also lighten a bit but what I like most is the shape of the rocks, mix of bumps and fractures. I don't like that mesh out of bounds because it is not a clean way to work, but being those overhangs of the tile unwalkable, it is doable '>

 

looking forward if your choice with terrains and raise height will allow to create a good mix of steepness and gentle hills. I challenge you to avoid the pyramid and make something different and better than lord rosenkrantz rocky mountain tileset!

               
               

               
            

[Legacy_MerricksDad]

               

I like the texture although I would cut some purple and leave space to more gray, also lighten a bit..

 

It is gray. The purple comes from the blue light in the area. As mentioned, I'm toning the texture for use with default area environments. Here's the base texture without any tinting from light or environment:

 

 

The red displays the quantity of feldspar in the rock, where the gray displays the quartz and mica. The speckles of blue can represent crystal faces, as seen in many places within Custer State Park. Still not sure I like the moss being green. It should be blue and rust lichen.

 

I made this texture by combining about 20 images, one of which was a hand-painted stone texture from TL2, which gives me the base darker lines to work with.

 

A previous version of these tiles had an on-tile tinting of a bit of yellow. Torchlight really picked up the color and made it look more realistic, just as the sun does on The Needles near Sylvan Lake. The quantity of feldspar really shines then, and most of the rock becomes yellow to brown.

 

 

looking forward if your choice with terrains and raise height will allow to create a good mix of steepness and gentle hills. I challenge you to avoid the pyramid and make something different and better than lord rosenkrantz rocky mountain tileset!

 

Guessing you haven't actually downloaded the dropbox file to see the variant heights yet '> Wait til you see the finished 20m cliff in the next release. Spectacular! I'll be able to easily recreate Harney Peak, just like I've wanted since 2004.