[F8] Orthographic engine with node based pathfinding

[renderhjs]

current object inspector debugger:

shows me the current registered objects, their icons (if any) and their actions)

smal update:

3d statues (mapped the previous images on 3d- meshes)

exit icons & notifycations:

icons on pickup- able objects (the one you can store later)


http://www.burg-halle.de/~st2520/

[trogdor458]

this seems to be becoming more than what it was meant too
it's like a coffee lounge where you can relax, no baddies to shoot at
but there is no one there

this seems really familiar, you have to do stuff to get somewhere
what a shame, i used to be able to view this place, now it wont let me anymore (added it to my trusted site too)

[Captain_404]

The updated version isn't working for me...it's just the previous version. Perhaps my computer is just messed up.

[renderhjs]

stupid me forgot to upload, then uploaded the wrong thing,-... will be online soon

[Captain_404]

yep, it seems updated now, I don't know if your supposed to be able to pick up items in this version, or look at the ones you currently have, but I can do neither.

The angel statues look a lot better now though.

EDIT: I think you may have replaced a lamp graphic with a coca-cola can. The art-deco room now looks like m. night shyamalan got sponsored by coca cola...

[renderhjs]

no inventory yet,- that´s next up. And I need to add a speech/baloon engine to it to actually let the player talk or say something (some actions will trigger that). The inventory will be pretty solid though I need to store special actions in the levels in the shared object of flash to reproduce with each new loading of what some did before in that room (e.g picked up the can, so let that one disappear,...)

[Evaman]

My friend has a question, what version of 3DS Max are you using?

[renderhjs]

r8 but honestly all the iso- rendering stuff can be done with max 2.5+ as well. Just the texture baking was introduced in the version 6 or 7,
why he want´s to know?

[Evaman]

He uses 3DS Max, along with myself very often. He's currently using Max 9 the, latest. I showed him the link to your game, and he is very interested in how your making the whole thing work with models using XML scripting.

[george_margaris]

that is a most fantastic thread.

[austriaman]

(...) I need to add a speech/baloon engine to it to actually let the player talk or say something (some actions will trigger that). (...)

FANTASTIC work. You should consider introducing sound.

[chrimarg]

in the future they will hand out little renderhjs, instead of sdk`s, you just have to tell them about your gameidea, feed them and give`em shiny things.

great work fella, Im utterly impressed by your drive and the results.

C.M.

[renderhjs]

thought I´d post it here as well,- been working again on this project,- been working on this knight the last few days. He´s not yet in the right pose,- but he´s supposed to pose next to doors and in hallways with a sword in his hands.

once he´s done I setup again a orthographic camera and render him from 16 - 24 angles so he turns smoothly in my engine

[mr_malee]

wow, nice work, that looks amazing

i would love to see that in the engine.

[renderhjs]

thx malee

almost done (just leather and other little details left out):

the knight texture is 4k (4096²) and the sword 2k. At this dimension some spots get blurry because of the texture size - guess that´s about the limit I can render with this texture without blurry details.

[T1ger]

Just... Totally amazing.. I have no other words..!

[khaine666]

How on earth do you build/design this stuff? It looks absolutely stunning!

You seem to combine design with programming just fine, wheras most people aren't even half skilled in either one as you are. Keep this up!

/me sticks two thumbs up!

[rachil0]

You mentioned early on the importance of fast line segment intersection tests because they get called so often. I thought I could share another way to compute it which might be of some use - maybe it's faster (but maybe not who knows).

Figure 1 shows line segments (A,B) and (C,D). Start off by computing vectors AB and, AC and AD (6 subtracts). Then compute z1 = AC cross AB and z2 = AD cross AB (4 mults and 2 subtracts). In 2D these are just scalar quantities that represent the magnitude of a z-componnent "out of the page". AB cuts the page into two halfspaces, the sign of these cross products indicate which halfspace the points C and D lie within. For (A,B) to intersect (C,D) the signs of z1 and z2 MUST be different - C and D "straddle" the infinite line passing through (A,B).

If the signs of z1 and z2 are the same, return "no intersection". Otherwise...

(Figure 2) Repeat the check using CD as the cutting line, let z3 = CB cross CD and z4 = CA cross CD. Note that for z4, you already have CA computed, it's -AC, reuse it by computing z4' = AC cross CD, and z4= -z4' ). To get all this done takes 4 subtracts (to make CD and CB - we don't make CA), then 4 mults and 2 subtracts to do the cross products.

Again, for the line segments (A,B) and (C,D) to intersect, the signs of z3 and z4 must be different. If they are different, return "yes, intersection".

Figure 3 shows a non-intersecting case - C and D lie to the same side of AB, so the sign of z1 = sign of z2. This is the fast failure case.

Figure 4 shows another non-intersecting case. C and D straddle AB, but A and B don't straddle CD. This is the slower failure case.

Figure 5 shows a degenerate case - all four points are collinear to within machine precision. When these kinds of cases pop up, your cross products will be extemely small in size (like less than 10^-6 or so) and their signs shouldn't be trusted as accurate. Bounding boxes can still bail you out for cases like this.

There are additional degeneracies where the lines overlap and intersect at infinitely many points, or intersect right at their endpoints. These are all characterized by (near) zero cross products popping out - it's tough to say what correct behaviour is in these cases, that's more up to you as the application designer. For these degeneracies I would just flag the result as uncertain (through the return value) and then leave it up to the caller to figure out what to do (usually this means calling another, more certain, checking function).

As an implementation remark, I would do this all "in-place" just using doubles allocated off the stack (variables local to the function) instead of new-ing a bunch of vector2D classes and calling .cross and .dot member functions. As a line intersection predicate, I would think this is generally cheaper than computing the intersection (which takes 2 vector subtracts, 2 normalizes, 3 vector dots and various scalar subs/mults/divides - at least the way I do it. It's those normalizes that hurt the most, but maybe you have a faster way). Of course if you actually NEED to compute that intersection and use it later, this tidbit might be of limited use.

[Cimmerian]

Hey render, how much time did you take to reach this level of perfection?
And, is there any book/online tutorials about 3d max that you recomend the most?

[walsher]

I love the model. I guess when I think of armor like that, I think more shinny and polished. But the textures are well done. Wish I could remember how to unwrap. What dimesions are those maps?