649 réponses à ce sujet

[MerricksDad]

Definitely planned some ledges. I like em like the ones in Dungeon Siege, just as you suggested, they are small parts of the tile, and have walkable locations.

 

I had previously done a ledge system with a module of mine back in 2006-8. One of the characters had a set of teleporation boots, which were very short range. He could jump from ledge to ledge, or in the case of the chasm, from the top of a pillar to the next one. We kinda ran the game like a Legend of Zelda adventure. Certain items let you get to places within already explored areas. Or go to areas previously inaccessible, but that you had passed by hundreds of times before, just knowing there must be a way to get that item you could see.

 

And I really like the goat idea! I had not thought of that. Now I want to put invisible strips of walk space on the rounds of my granite faces. Just like in some of my Harney Peak photos. Can't believe I didn't think of that

 

I really like when multiple brains meet in a single location.

 

I think from now on, I may show screen shots for this area in the guise of a mountain goat.

[MerricksDad]

Didn't know if you guys wanted to see a close up of the current texture. This is more seamless and slightly lighter than the one in the previous shots. I think it will go better with the higher quality grasses I have (but not using yet). Again, this is a 512 image, and this is displayed at the maximum settings on my crappy dell laptop with old intel chipset. Shadows are screwy still, but lighting is great.

 

[Wall3T]

So i was watching alot of the progress and someone brought up the idea of a Highland Tileset, which reminded me i made one for my server a few months ago using a retexture of mondegos mountains (which is a retexture to Lord Rosenkratz Tileset).

 

in mine i couldnt quite get the rocks to appear as seamless, but i modeled mine after the Scottish Highlands to be designed in that purpose. i thought it intresting how close it was to yours and wondered if you were taking a similar approach?

 

im not much a tileset designer so im limited to what i can work with, but itd be cool to have something similar to what i have below

 

[/URL]">http://

[MerricksDad]

I see that one has additional grass-things added, as will mine. Other than a few texture issues, as you mentioned, that looks like a great set. Right now, I am using a 1001cm cube to texturize anything on the tile. This is generally insufficient, but as you can see with mine, it does the job for now. I then rotate the cube 45 degrees on the Z axis so I can force more of the part to see the texture, and it also mixes at points other than a plain cube will. A plain cube will highlight things like 45 degree angles, and the edges of your tile which face those specific directions, and that only makes your tile look blocky and shows of the squares.

 

So I just finished the basics for shallow lakes. Again, they are not properly textured, but I did get the a-nodes set up correctly, and have used my tx_wateranim08 from that pack of Torchlight 2 water textures. It looks really good right now, and will look much better when variety, decals, and vegetation is applied.

 

And! Just in case anybody wants to play with it now, here's a dropbox link. I will continually update the file, so expect it to grow each time you download it. 16 megs right now. You can go in and at least check out the contour for yourself. Feel free to play with the fluff and carve as much as I did today. And don't forget to check out the album I'm building. Link should be on the first post.

[rjshae]

I have a curiosity question: can a layered mesh with an alpha-transparent lichen/moss texture be used to break up that rock  texture tiling? Or would that bog down the performance too much?

[MerricksDad]

I plan on it yes.

 

And no, small hits here and there won't damage the play speed. Hell, if I can use a 1024x2048 cycling animated texture for the water without anything but a load-time hiccup, then I can still do just about anything else I want with basic decals. And if mine can handle it, I am certain everybody else can handle it. That is one of the beauties of building everything with an ancient device

 

Vanilla used a lot of transparent textures. The key problem with transparency is when you cover a large portion of the screen with it. What I plan on doing is covering a larger portion of the base texture with something else, like vegetation. And keep in mind, you won't be able to go examine large portions of the rocky masses, so you won't ever get to see the texture lines in game. It won't be naked as shown in the images so far, but fully dressed with a high polygon count. The trick is making low poly shadows for everything, and breaking the key features into facing groups which don't interfere with each others shadows.

 

Take for instance the line between mountain bases and ground bases. Right now, the entire section is a single model, with a horrible shadow issue. It lights wonderfully, almost perfectly. But it's only one piece, and that causes a lot of issues later. What what you do is get a non-transparent texture to straddle that line, which creates two more lines. Then you ride those lines with a very thin strip of transparent decal specific to the transition, reducing the overall calculation of transparency. And then, you cover anything unsightly with a sub feature, like a plant, tree foliage, rock pile, etc.

 

Same goes for water/beach transitions. Three basic textures, one or two transition strips, and then details.

 

Keeping those transparent transition strips small is key. Even though a large portion of the image may not be transparent, the larger the image is, the more calculation will be done. I shouldn't say the larger it is, I should say the larger the polygon using it. That is more accurate, because it isn't how much you are mixing, it is how much you are mixing it with, and at what levels/distances. If you constantly have to blend large regions in front of the camera, from all the way back, all the way up to the camera, you are doing a lot of calculation in a single screen draw cycle.

 

What your output will actually be when this is all finished, is a completely seamless landscape, very much like that you can create with NWN2, just without the bump mapping....well for me anyway, since I can't do bump mapping on this computer. Bah.

[MerricksDad]

I currently built the mountain terrain different than I built the 2m raised grass sections. Those grass sections have model parts which exit the basic tile size. This was intentional. What it does is break up the seams on a tile and carry the "crack" in the rock over to the next tile. You don't know where it will land, because the shapes spread into the next tile by about 1/16th. What it meets there is how it is drawn. When each tile has a slightly different escaping shape, what you get is "randomized" seams, unless you repeat the same structure over and over again using the same tiles.

 

If you take a look at the mountains, I've left them as most people do, and that gives me the current ability to use the same shape as a temporary walk mesh. Again, that is just temporary. I still need to detail out every single tile. That includes increasing the poly count of the shape you see right now by extruding selected sections of the cliff face.

 

Take a look at this image. This is what it would be like right now if I called it done, and updated the textures like I did in 2008. (not my image)

 

 

Now, take a look at this one, which represents how I intend to extrude the visible mesh BEFORE adding decals:

 

 

You can see the second image would have a lot more polygons, more shadow to render (to an extent, it can be faked), but would take a texture better when each individual extrusion was properly wrapped.

 

Keep in mind that I intend to wrap certain parts of the extruded areas into the next tilespace. This will completely get rid of all texture lines, except where I use decals instead of extrusions.

 

So then, when that is done, I simply take the cloned walkmesh from the original shape, remove walkable from the section no longer walkable, and then reduce the walkmesh polycount by collapsing off the rounded faces of the cliff face. Alternately the current walkmesh could retain its shape for a better collision system for the mouse pointer, but it is not required.

[MerricksDad]

So what Torchlight does, and I am sure many other games do, is they take those extrude sections and give them their own chunk of texture. They also give those extruded sections a separate shadow, so when environment shadow is turned on, it properly affects those parts behind or under itself. Single mesh landscape tiles do not properly affect their own mesh with shadow. The engine has a very hard time doing that, and makes a terrible mess of it. Sometimes shadows are too dark, sometimes they have big gaping holes in them where you know light should not come through. And sometimes, a simple corner which has a different smoothing group simply does not render a shadow at all, and without any reason.

 

By extruding those sections away from the main mesh, that also creates more seams, but since each extrusion finishes its texture mapping before leaving the mesh edge, it doesn't matter. You simply make the edge of the extrusion equal to the part you won't be examining. Some people have previously failed in this, including myself, by not separately wrapping the extruded areas. Instead, they/I have tried to apply a texture to the whole thing after the extrusion, and that only causes more seam lines to appear within each individual mesh. You really only want the seam lines to appear at the most useless parts of any mesh. With rocks or rocky outcrops on cliff faces, that needs to be either at, or behind, the most primitive face of the cliff. In this case, the primitive face is the shape you see now.

 

I will edit these posts and add some example pictures of what I mean when I actually finish a single tile. I might do a tile or two this week just to show what the final product will look like (minus the vegetation because it isn't completely grown yet).

[MerricksDad]

Here is an example of a Torchlight 2 texture which is not meant to be wrapped around a structure in the normal manner. Instead you extrude faces which match the texture, and then stretch the texture around the extruded parts in such a way that all the edges match up. I think they actually hand painted these AFTER making the model shape, rather than what I am doing.

 

 

So that is all done with a single texture. But take a look at this next one. This is a separate texture for extruded faces, or placeable rocks.

 

\

 

This is set up more like how we do for creature parts (which on creatures I am not very good at doing). Again, you can see that the texture is built in such as way that it wraps an outcrop feature, but also has lines which can be used to break up that feature futher via poly division. The seams of the texture wrap behind the visible part of the object, and you just push this wrapped object into the previous, and it looks good.

 

Look at this mess!

 

 

But you can see that it is actually a texture for multiple objects, all bundled into a single plate. Each section has a darker line around it, which is intended to butt up against another part, or object, or be hidden from view.

 

Edit:

 

Another very important part about using textures this way, is that you can use different angles across the texture. Take for instance the third texture above. Parts are made so that they can wrap around something. Normally we build a texture for something by using a skinner to disconnect all the faces, like we do with paper dolls. Then we color those faces to fit our needs. The end product still uses linear sections of that texture. What I mean is that all the grain marks on that portion of the texture are going the same direction. In the texture above, the way they wrap stuff so that not every vertex is equally spaced in the mesh as it is in the texture, allows for some distortion being added to the texture. This causes the grain to vary in size, shape, and direction over the faces being supplied for.

 

This is how I intend to texture the extruded parts of the mountain regions. the rest is just basic shapes.

[MerricksDad]

Here's an example of how much of the base texture seams can be covered up with just the tileset-created grass. Playing "spot the looney" here. Can you find him?

 

 

And here's an image of the initial water texture, showing 50% opacity, with a custom lake bed texture underneath. The lake bed shows only a tiny quantity of detail through the water surface, and you really have to get close to see it.

 

[Wall3T]

very cool stuff im about to go look at your set MerricksDad.

 

in my original plan i wanted a Skyrim-ish planescape with the moss on the rocks, but i ran into the tileing problem. the moss looked great, but i found it just didnt look right to me. Such as every rock had moss on it, which seemed unrealistic to me.

[MerricksDad]

Sure, I completely understand. When you simply repaint a texture, that part you painted affects all the models which use that texture. What you need is some decals. Then you clone the part of the rock you want to apply the decal to. Make it so the decal fits the cloned portion correctly, having enough transparent space to make up for anything you don't want textured. Then make sure the child order sets the decal later in the draw indexing, so it always draws after the part under it (otherwise it flickers). If it still flickers, you just alter the decal by positioning it a fraction of a point away from the original. Of course all that requires actual modelling, and if you have a lot of parts to apply decals to, then you are looking at hours of work.

 

 

 

Tonight I finished the prototype for deeper water. I'm not completely happy with it, same as with the shallows, but the water alpha makes the deeper water look cooler. If you hit it just right with the sunshine, it looks like the centers of the lakes are shinier. I think it's the fog setting doing it. Water attains the fog values down into its depth, lightening it instead of darkening it. Something to work with later.

 

Now I gotta figure out if I want to do more water tile variety first, or move into stream prototypes. Also, do I want to do shallows mix to water, or do I want to do extended beach first? Seriously trying to get the prototypes out of the way first before I go crazy on the details.

 

Edit: oops, just noticed the sword in my previous image does not show through the water. Gotta figure out the magic key to fix that before too long.

[Frith5]

Just took a quick tour of your prototype. Outstanding so far. The water, to me, looks great. But what really got me was the closeness of the defiles and ravines possible, what will everything walkable atm. The openness of NWN, although I understand the reasons, has always left me a bit unhappy. The rocky slopes crowding my avatar were oppressively nice, if that's possible. I'm eager for more.

[MerricksDad]

I think I found a way to at least make nearby water look deeper as it gets deeper. Still lightens up due to fog at a distance.

 

[Killmonger]

 

Wow md !

 

Really nice idea/ project /progress !!

 

Will there be vertical cliffs (promontories) too?

 

(I could envision a few scenes from The 300)

 

[MerricksDad]

Not yet finished, but if you combine the mountain (raise/lower) with the 1m or 2m grass rise, you get a more steep cliff of 1100 or 1200 height. I am thinking of bringing back the half height corner too, because I realize I actually do need it for talus slopes and glaciers. In the previous iteration, I had used all half heights, quarter heights, and double heights to make the 8 height transitions. (quarter, half, three quarter, full, full plus quarter, full plus half, full plus three quarter, and double). I don't think I will do a double height corner, as sheer cliffs of 2000 height over a 1000 wide tile looks kinda weird, and makes for a really difficult cut away box in some scenarios.

 

As for height transitions into water, definitely yes. I want to do 1m, 2m, and mountain directly into shallows or deep water, just like they did in Dungeon Siege. They pulled off both sharp and smooth slopes into the water, and it was nearly flawless, give or take their low quality textures. If you look up Sylvan Lake South Dakota on google maps, you can see the rock-to-reservoir images I will be basing those transitions from. At the back of the lake they have mountain to water, and at the one side of the lake they have half height to water.

 

[MerricksDad]

Thinking about changing the fluff crosser to "Dilate" and creating an "Erode" crosser similar to what I already built for the original carve (before this sharpening carve). What I'd like to do is make is so if a "Dilate" crosser is used, the walkpath is completely closed on that portion of the tile, at the bottom, not the top. The top gets more space to walk on. The walkpath on the adjacent tile will still be open, but that makes the area you can walk in half as wide. Now if you use the "Erode" crosser on the same sections on the adjacent tile, then that walkpath on the top will be closed, making more room at the bottom of the tile. When you combine them, it makes a 200 meter drop, with no walkable ledge at the half way point, crossing two tiles. This could be how we can make taller mountains without making double high tiles. And it gets rid of another of the texture seams.

 

I'm exceptionally busy this afternoon, but I might be able to pull this off late tonight, since I don't have to work tomorrow.

 

[MerricksDad]

And here it is: fully compatible Erode/Dilate crossers, creating a 2 height transition mountain in approximately one square.

 

 

 

Seam is minor, but gain, this is the primitives stage. Both dilate and erode crossers in their primitive form have huge shadow issues, and also create a lot of tile boundary edge visibility. This is because they take right off from the first edge on the tile, rather than the second edge, as with all previously shown tiles.

 

In my first tutorial on tileset building, I stressed the importance of using at least a tiny portion of the tile edge to make a seamless area. That area needed to duplicate the exact position and face angle as the adjacent tile. This would virtually erase all face to face smoothing issues, as well as reduce nearly entirely any glares caused by those edges. I didn't follow that advice on these two crossers.

 

It is possible that I might go back and manually edit the edge smoothing on these, but I don't think that will be needed, since i don't plan to keep the primitive seams anyway. Additionally, it would take longer to manually manipulate all those verts on 30 some tiles, than it would to redo the entire set, which is what needs to be done when I extrude and detail the surfaces anyway. So ignore it for now. I will.

[MerricksDad]

Improved lake and deep water edges. Now with mixing of the two, complete with shoals. I intend to add stacks, arches, and rock piles to these in the future. Deep water is not walkable at 5 meters. Shallow water is walkable at 1 meter. You can see the completed deepwater texture, created dynamically using a gradient overlay. It carries across all water tiles now, using the proper depth for all tiles.

 

[MerricksDad]

Spent a bit of time doing decals and boulders today. I found a boulder shape and texture which suits the needs of the tileset, and have three decals of rock piles ready. Grabbing a few ideas from TL2 and mixing them with NWN.

 

[Tarot Redhand]

Quick question with a yes no answer. Are you going to do some scree?

 

TR

[MerricksDad]

Yes

Definitely! Talus slopes are next on the to-do list, actually. They are my favorite place to rock hound! But as the original Legend of Zelda shows, you gotta watch out for falling rocks! ...Did anybody other than me think that those bouncing boulders in the original looked like giant boogers?

[MerricksDad]

After four versions, I decided that talus slopes might be best as an alternative to erode crosser. I may make scree piles a separate terrain at some point. Trying to make approximately 35 degree slopes up against 45 degree slopes is difficult, so I used the sharper slopes from the eroded set, with nearly maximized slopes. This also makes it look good when used with the dilate crosser. Not so much with the fluff crosser, because of the primitive ledges. Right now the walk mesh still hugs the original rock structure, but scree will be walkable in the final.

 

 

[MerricksDad]

Now that we've got a few flowing water textures, and have some more people starting to play with animesh, next on the agenda is directional streams and rivers. Yes both. I want to bring back my 2006-08 directional streams, and also make crosser-style rivers. This will be complete with basic waterfalls. More waterfalls will come later. I just want the prototypes out of the way first. I'll very likely do waterfalls to cross all my mountain-styling crossers. That added variety will make some great future cave entrances.

 

As I did with my streams before, I'll do right side and left side streams. Rivers will go up the centers, as most people use crossers. This will allow both right and left handed streams to merge easily to slightly larger rivers. Since they're directional, they'll merge from the side at acute angles, or possibly have some interesting groups to merge those various waters.

 

I won't make it now, but somebody may later: since I'll be doing directional flowing streams, you might want a backward flowing waterfall, like the one they tried to use in the DND cartoon.

[MerricksDad]

First look at the mountain streams.

 

I have a few issues with the moving water, but I'm getting there. The big issue is either 1) wrapping the texture flawlessly, or 2) using enough "stuff" to hide seams. This is only a problem when the streams split, merge, or widen.

 

In the image below, I've made a test split. This uses three sides of the tile with stream crossers.

 

 

I'm using the TL2 extracted water. This one is the tx_wateranim04, with TXI, as packed in that kit I put on the new vault. I've tiled it 4x4 to give it more detail per meter. This is 75% opacity. I may increase the frame rate to speed up the base water, but mesh stretching should supply the rest of the wibbly wobbly stuff I need to make it look real a it passes over rocks, or swirls at intersections.