Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Starver]

(While I slept overnight, BTW, I recalled an old 3d-mesh terrain trick to stop edges of Level Of Detail look too straight and artificial as you look out upon a vista of procedurally-tiled landscape with limited processing power to render. No matter what your tiling method for chunk data - usually square for easy arraying but perhaps triangular, possibly hexagonal and, in some cases even already unregular convex polygons - the LOD definition included a serated (essentially concave) bite out of each straight edge that is designed to create a new (convex) area across the tile border with whatever neighhouring tile (and facing 'litoral zone') it ends up having.  Then when deciding how to display the two tiles (perhaps with a different LOD due to a visual distance element, but also helps between 'biome' transition, in DF term) it fills the interiors (less border-bites) with the stock info based on each tile alone and then grabs and uses the appropriate 'litoral' bite-filler to smooth the transition based on both the neighbour-tiles involved*, for absolutely every set of defined tiles involved. Now, this is 30+ years ago in game design, and intended for 3d, so not a suggestion to try here. That said, if (e.g.) a "corner of road tile" had a chunk out of its 'road end' sides that could be swapped over by a shared border (sub)tile if the next tile jinks the road back again, it could like like a different transition to if the next tile takes the road straight, or corners back on the first tile's entry, or stops at a settlement graphic (overlay). And so N types of tile that are painted up get supplemented with up to N² little 'edge diamonds' (though each with <¼ the pixel data needing artistry than a main tile) to at least cover a sufficient sample of transitions and keep them from looking too much like the tiling that they are. I'd do a graphic mock-up, but my artistry is sub-par and this was supposed to be a short, idle note, not a rambling screed like it now is. And that having derived from half-awake musings either side of a sleep-cycle, so probably not entirely useful to real-world applications any more.)

TL;DR; - I have remembered some method that's a lot more work to set up to then make an infinite number of procgen maps look somewhat more hand-tuned. And it probably won't delay every other feature by more than a few months, right?

((* The bites angle into the corners, meeting other bites only at multi-tile meeting zones at the corner so are designed to transition point 'stellated' own-tile-only next to two-tile-dependent back to next-tile-only and so on, though a more complex version also added corner-biting options that required a far more massive library of transitions to cover every possible 'circuit' around a corner, not just tiletypes² number of edges, and that could be tuned and exhaustively tested in every possible combination to produce an array of not-obviously-a-corner-joint corners.))

[therahedwig]

There's something slightly unreadable about the maps. I personally thought it might be that each tile has a little bit too much brightness detail, but when I went to desaturate it to prove, I noticed there's also too little variation in brightness between different areas.


So topleft is DF, topright is dragon quest 6, bottomleft lufia, bottomright breath of fire 2. You can clearly tell the forests from the mountains from the plains even in the simplest (lufia) map through sheer brightness.

Most importantly, even when greyscaled, the sites in the other games' maps are still pretty easy to pick out, especially compared to those forest retreats in the DF map.

From an accessibility standpoint you should proly desaturate your images from time to time to see if they're still readable, as the perceptual brightness is the most important color contrast. You should prefer a gray filled layer blended with the color blending mode over the desaturate filter in most graphics programs. (as you want a luma/luminosity style calculation instead of a flat brightness calculation, as the luma calculation recognises that yellow is perceptually brighter than red, and the color blending mode as per pdf standards uses a luma calculation, hey, can you guess my dayjob)

Like, I know it's not entirely fair to compare DF maps with old snes maps, as those had far less detail than the DF ones do, but I think because DF has much more detail of it's own, it's pretty important to have the most important features highlighted more?

Do you guys plan to have seperate tiles for the adventure mode travel map (as in, there's more zoomed in variations when you travel in adventure mode, a mid-tile one, and an even more zoomed in one for the sites, like the cities). Also, what is that yellow thing above the hamlets? A keep?

[FrankVill]

For me the greatest achievement is the readability of the map at a glance. It's great to know what each thing is (without having to go to the wiki )
However, all the mountains seem the same to me when they are not, there are high mountains that end in a peak and medium mountains whose summit is rounded, but you have to get close and look closely to notice it. Possibly not a priority, but I think it would be nice to be able to distinguish them a little better. It occurs to me that instead of adding more detail to distinguish them you could alter the color slightly (for example, light gray for medium mountains and a little darker for high mountains).

On the other hand, what I will say next should not necessarily be considered a criticism, it is only about my impressions due to my personal tastes. The colors seem too bright to me and some shade is missing in some figures. For the devilish areas I would have chosen a darker purple color or one that feels more threatening. Altogether, everything transmits to me the feeling of a very happy world, when the truth is that DF is full of depressive dwarfs, haha.
But even with all that, it seems like a great job. The whole set feels well integrated and single-bodied.

[Manveru Taurënér]

Also, what are the yellow-whitish little things on top of the hamlet(?) sites supposed to be?

[mko]

Also, what are the yellow-whitish little things on top of the hamlet(?) sites supposed to be?

Fields?

[Starver]

By the way, bearing in mind my own given my lack of personal artisticness, here's an experiment I was making with the Windmill.

Spoiler (click to show/hide)

Right hand-side is a representative of the side-view laid down.
Yes, different speeds[1]/directions (and updating frequencies, to lessen the number of 'frames' involved[2]) and the occlusion of the lower blades (marked darker on the two side-plans where they are so marked) makes things messy even though I adjusted the imagery dark for lower vanes (maybe should have been a different adjustment, or completely mask the transparent gaps?).

What I learnt from this is that "it's not easy to do". Not without going for something different, perhaps needing entirely new art to work with instead of what I ripped from that original X/+ two-cycle-per-quarter-turn model.


Windmills finished with, from my perspective. Definitely. Until next time!


[1] Original animation used 0.7s per quarter, with a small timing error at the start, and I fudged it again a bit to match different 10ths of seconds cycles for 3- and 4-blade cycles converging at 4.8s rather than 4.9s, in the four blade and three-blade versions. in my head, there was always a log 'support' and then either two crossed long-logs and a power-log parallel to support, or a brace-log at a back-angle, or it's one support log with three blade-logs, so I was playing around with that, but of course have no skill of artistry I just took what was there. And with a vertical post to get the vanes past I had to shift the 'arms' forward.

[2] For demonstration purposes, I could have done it smoother, but ultimately if the game were to go far beyond a 3-cycle of animation it'd be putting a load on both engine and artist to provide that. However good it looks, it wouldn't be worth it. Though I think I've proven to myself that a 3-cycle (of however many blades, not including 2-blade which would likely be worse) doesn't cut it although a 4-cycle of three blades might work if it doesn't seem to shimmy even worse.

[Manveru Taurënér]

Fields?

Oh, that'd make sense I guess, not that I'd have guessed it. Looks like a weird building of some sort, doors and windows and all (how it partially obscures trees behind them doesn't help either if so).

[jecowa]

I'm trying to figure out how this works. Even though a map tile is 16x16, it looks like some of them are larger. For the parts of the tiles outside the tile's normal boundaries, it looks like large mountains have priority over small mountains, small mountains have priority over trees. It looks like large mountains always cover the small mountain, and the small mountain always covers the surrounding terrain. For tiles that cover surrounding tiles, it looks like they use 3 layers to determine what is drawn on top of what.

It looks like there's multiple tree tiles it can randomly pick from, but it's difficult to determine how many tree tiles it can pick from because adjacent trees tiles are drawing more trees on top of each other and it looks like there's more tree tile variations than there actually are. Maybe there are 4 tree tiles?

[Meph]

Thanks for all the feedback so far

I'll go over it today and see if I can improve the map some more. It's only the first iteration, so be patient. It's work in progress.

Jecowa, it's 5 forest sprites, but sometimes multiples are copied into each other, offset a bit if the neighbouring tile is also the same forest.

[Bumber]

-snip-

How about 4 blades, 4 frames?

[Starver]

0(+90+180+270), 22.5(+...), 45(+...), 67.5(+..) leading back to 90(+180+270+0) might work per blade, but in an early test I was very unhappy with the ±22.5° from straight up phase hidden in the 67.5° clutter behind it that had to be drawn. The vertical blade in the existing + format of the 2-cycle is wide[1] at about 14px (IIRC) around centre-zero and the 'face' of the next frame (uncomplicated by rake) was from something like 2 to 12 offset pixels, thus narrower, and yet must look feasibly like the 38ish pixel length of a completely outstretched blade of the ur-example I was pinching from[2].

And redrawing as William Cubbit-style shutter-sails (with or without self-regulating, or 'spider'-controlled shuttering controlled from behind the spindle) is a bit off-era, even if I could do it.

Windmills finished with, from my perspective. Definitely. Until next time!

Until next next time...


Darn, deleted my footnotes, prior to preview, so what I had tapped in was...
[1] ...something about the rake angle as it passes over the top. (Reverse rake at the bottom is hidden, of course, though could stick out.)
[2] ...something about the outstretched spar having 3px of strut, 2px of gap. Like a line of fenceposts seen from a highly oblique angle the gaps disappear into and the struts overlap, to the tune of 1¼px where there were 5 of already pixel-width detailing. Beyond my skills to redraw, and doesn't work well with digital rescaling in lieu of any such skill.

[Meph]

0(+90+180+270), 22.5(+...), 45(+...), 67.5(+..) leading back to 90(+180+270+0) might work per blade, but in an early test I was very unhappy with the ±22.5° from straight up phase hidden in the 67.5° clutter behind it that had to be drawn. The vertical blade in the existing + format of the 2-cycle is wide[1] at about 14px (IIRC) around centre-zero and the 'face' of the next frame (uncomplicated by rake) was from something like 2 to 12 offset pixels, thus narrower, and yet must look feasibly like the 38ish pixel length of a completely outstretched blade of the ur-example I was pinching from[2].

And redrawing as William Cubbit-style shutter-sails (with or without self-regulating, or 'spider'-controlled shuttering controlled from behind the spindle) is a bit off-era, even if I could do it.

Windmills finished with, from my perspective. Definitely. Until next time!

Until next next time...


Darn, deleted my footnotes, prior to preview, so what I had tapped in was...
[1] ...something about the rake angle as it passes over the top. (Reverse rake at the bottom is hidden, of course, though could stick out.)
[2] ...something about the outstretched spar having 3px of strut, 2px of gap. Like a line of fenceposts seen from a highly oblique angle the gaps disappear into and the struts overlap, to the tune of 1¼px where there were 5 of already pixel-width detailing. Beyond my skills to redraw, and doesn't work well with digital rescaling in lieu of any such skill.

By this point you are just stringing up made up words. XD

[Starver]

Arglebargle?

[jecowa]

From watching the vanilla windmill animation, I get the impression it moves something like this would:



Not a screw design, though, but something that would spin in this manner.

Edit: something more like this:

[bloop_bleep]

That design wouldn't work, though, since the wind would turn the panes orthogonal to its direction and then afterwards blow equally hard on all the panes. You need a tilt or aerofoil.

In fact the screw might actually work a lot better.