Incline SHELTER update thread

[soggie]

Now my only problem is how to mix 2D backgrounds with 3D characters, and then how to use pixel shaders to render real-time lighting.

Here's the general idea:

1. All art assets must be rendered with: albido (diffuse) map, normal map (in world space) and depth (bump) map (also in world space).
2. Render pass 1: render all albido into a render target
3. Render pass 2: render all normal maps in world space
4. Depth sort all entities with Y-axis and render each sprite's depth representation (no fucking clue what this means)
5. For each light source use albido, normal and depth map to calculate light values per-pixel

I'm still trying to figure out what all these means to be honest. The last time I dealt with all these 3d jargons was 6 years back.

 

[shihonage]

This is way above my paygrade.

 

[Prosper]

you can use a series of ray intersection tests to calculate the xyz coords and then plot the pixels as 3d points using those coords.

 

[soggie]

you can use a series of ray intersection tests to calculate the xyz coords and then plot the pixels as 3d points using those coords.

Wouldn't that that be insanely expensive, cycles wise?

 

[shihonage]

Thinking small isn't Prosper's style.

 

[Prosper]

you can use a series of ray intersection tests to calculate the xyz coords and then plot the pixels as 3d points using those coords.

Wouldn't that that be insanely expensive, cycles wise?

you only will need the amount of pixels of the actual object to be represented. the ray test is done separately (once) and there on the game just load the data.

 

[shihonage]

soggie, actually the previous problem wasn't entirely resolved. I was breaking my head for the past 2 days trying to resolve it.

The objects on Y-1 behind the multi-tile object were still incorrectly rendered on top of it, due to their X+ later loop cycle.

I finally figured out a solution, wasted an entire day on finding a proper way to inject it into the renderer without breaking it.

Hint: it doesn't involve changing drawing loop order again.

 

[soggie]

soggie, actually the previous problem wasn't entirely resolved. I was breaking my head for the past 2 days trying to resolve it.

The objects on Y-1 behind the multi-tile object were still incorrectly rendered on top of it, due to their X+ later loop cycle.

I finally figured out a solution, wasted an entire day on finding a proper way to inject it into the renderer without breaking it.

Hint: it doesn't involve changing drawing loop order again.

I'm not really sure I know what you mean. Got some screenshots?

Right now since I've decided to go with the Ogre3d engine, I'm going to try my hands at pseudo-deferred rendering. Writing directly to the depth buffer and stuff. Not sure if it'll even work though.

 

[soggie]

I think you mean this?

How do you solve this? I'm using hardware acceleration so I'm using the depth buffer to cheat my way out of z-ordering, but I'm interested to know how it is supposed to work on a pure 2D engine.

 

[shihonage]

Yes. It took me a while to solve this, but I don't blit the bitmap at the grid's left upper corner anymore. I blit it upon reaching the right lower corner.

DING.

 

[soggie]

Yes. It took me a while to solve this, but I don't blit the bitmap at the grid's left upper corner anymore. I blit it upon reaching the right lower corner.

DING.

Hmm. Would this fall apart when you try to render non-rectangle shaped entities? I have a wall arch that is L-shaped, and another environmental prop that is T-shaped. I'm guessing in your editor you can specify which tile to render from for each individual entity?

 

[shihonage]

It falls apart on anything that is not a filled, rectangular block. Thankfully I can create a T-shape out of two rectangular blocks

 

[soggie]

It falls apart on anything that is not a filled, rectangular block. Thankfully I can create a T-shape out of two rectangular blocks

One note of curiosity: you draw bounding boxes for your wall sprites in your editor right? Are they just for show or do you use them to calculate height properties like judging whether a person is partially or completely covered behind the sprite?

 

[shihonage]

^^^ I don't draw anything in the editor. I do precalculate low-resolution bitmap masks to determine what is behind/over what.

____________________
Some shots from regression testing level:

 

[soggie]

Fark, how the hell did I miss this?

The screenshots look good. Glad to see you do regression tests. I've seen lots of "professional" coders here who don't even know what that means.

 

[shihonage]

This will be an unusual update.

______________

We couldn't help but notice that the small indie games are exploding. It is as almost if people are actually willing to buy games with CGA graphics as long as they are actually fun (or somewhat fun) to play!

Impossible.

Then we remembered that we nearly completed a game with highly primitive visuals, which also happened to be monstrously fun.

Gameplay-wise, it is feature-complete and tested. All unique monster types are there, including scary bosses, as well as script triggers, items, Hexen-like hub travel, saving, loading, replay recording, driveable vehicles, secret walls, particle and lighting effects, and even 2-player co-op and deathmatch (over LAN, OR, if you're lucky and have a good ping to your partner, over Internet). You can even save and load in co-op! And press F10 to see other player's view!

Unfortunately, it doesn't run well on Vista/7 (if at all), and the "campaign" was a series of disjointed maps with an impossible difficulty curve, with no definite narrative or ending.

Despite this, GamesDomain liked it, and so did SHMUPS.

____________

So, now we are making it work well on XP/Vista/7, with a proper campaign, and then will be selling it for really cheap. This game is result of hard work, mostly complete, and if we ever hope to finish something as ambitious as Shelter, we should be able to finish THIS.

 

[shihonage]

I made a decision.

After I am done with Monsterland, Shelter will ditch all FO:T asset dependence and rely solely on sprites obtained from Reiner's Tilesets.

They are inconsistent in animation state availability, and a lot of stuff will have to be compromised and/or replaced with Nintendo-era effects.

Nonetheless, nobody will confuse Shelter with FO:T anymore, it will be its own game, and I will be officially able to distribute assets with it(alpha demo et al) and eventually sell it.

Yes, the looks will take a significant dive, but honestly, I don't give a shit at this point.

Before my cousin left the project, we planned on making a ton of sprite-based compromises for the occasion when "a mysterious artist will do all this massive graphics work for us for an affordable sum", and now those compromises will be simply applied to Reiner's isometric sprites.

For instance, you will not see what weapon someone is carrying until they stop (there may be an icon above them showing it when they walk, though). Idle animations will be disabled for many. Hits will be portrayed just through blood splatter and pushing of the sprite back. Deaths will only be shown in 1 of the 8 directions. Running may end up looking like walking, sped up, and with a hopeful programmatic blur effect. I will be trying to replace as much missing graphics with effects as I can.

On the bright side, Reiner's got sprites of children! We all know what this means...

His color palette is all over the place, so I will be desaturating a whole lot, among other things.

All in all, I'm excited about this project again. It no longer seems like a dead end.

But I have to finish Monsterland first. Been putting a lot of work into that, for the same reasons that "Scars of War" guy, Naked Ninja, wants to release his card game first.

Monsterland supports 2-player campaign coop, and now I'm working on resource gathering/mining element, Dead Space-like combat upgrades, and refining various other interactions, such as better utilization of driveable tanks and an ability to repair them.

 

[villain of the story]

Couldn't you use 3D models over 2D tiles? I thought FIFE supported that, no? Working with 3D models is so much faster, easier and cheaper. Thousands of free resources on the net! I also find it strange that you, as someone making an ASCII game, would be so motivated by graphics alone. You could simply use shitty sprites and tiles made in MSPaint in a couple minutes to develop the game proper and show what the game is like to kindle interest and to find volunteers for art.

On the bright side, Reiner's got sprites of children! We all know what this means...

That I can role-play Mohammed (PBUH*)? YAY!!

*: Pedos be upon him.

 

[shihonage]

Shelter uses a proprietary engine, written entirely by me, which is not FIFE, and it does not support 3D models.

There are tools out there that can chip away the 3D models into batches of properly angled 2D isometric sprites automatically, making the difference in work required ... not large.

In isometrics, you need to have a lot more than just one-direction-facing sprite in order to have your game be playable and testable. Otherwise there are too many abstractions that can be mistaken for something else. In animation systems, specific frames are tied to specific actions, which also makes it highly desireable to actually have multiple frames of animation in each of the 8 directions.

So my later equivalent of "using shitty sprites and tiles" was "using Fallout: Tactics assets". That stage is already over. Shelter generated some measly interest, but finding reliable people who are skilled artists and are willing to put in absolutely MASSIVE amounts of work, in a timely manner and for an affordable pay (especially not knowing if the project will sell any copies at all), has not proven easy.

Mostly our team of art volunteers consisted of ... enthusiastic people.

But I can't blame this on them, either. There's a constant feedback loop is required, as IN THE MIDDLE of a project one does not know how many resources and of what kind will be required, exactly. I need to communicate with my ... sprites, for the lack of a better word. I did not do that very well, because I STILL don't know exactly what kind of sprites will be needed, because the game's equipment is not finalized.

Long story short, with my cousin gone from the project, I am alone, and I intend to use Reiner's tilesets to the best of my ability.

In the end, the best path for a starting indie game programmer is to replace as much graphics with code as possible. Monsterland is the best example of that, and I intend to do similarly with Shelter. It's much easier to alter code at will, than to make people redo their hard work over and over based on the ongoing nature of the project.

I really only need the sprites to be animated to walk in 8 directions, and Reiner has that. The rest can be cheated around, palette-swapped, etc.

That I can role-play Mohammed (PBUH*)? YAY!!

No, but I won't magically make the children immortal.

 

[Surf Solar]

's been more then a week now, we demand updates.

 

[shihonage]

I bounced to developing Monsterland because at the time I saw no way for Shelter to be commercially viable with FO:T graphics, and wanted a chance of earning $ that would allow me to dedicate more resources to Shelter.

Yet Reiner's tilesets, though horrible, are free, and there's light at the end of the tunnel. More light for Shelter than for Monsterland. It IS commercially viable.

But I can't keep bouncing between projects like this, so Monsterland must be finished first.

If I start making weekly updates, they'll all be about Monsterland, an ASCII realtime overhead view shooter with a few odd elements and co-op.

Though Shelter was so close to the exciting stage. I mean, if I managed to import the Reiner's stuff in there, I could pretty soon release a Junktown alpha demo - one that would be infested with exploits, however. But, shelves would be lootable, people would be killable, quests would be completable, travel between cities would be... travelable, saving and loading also works...

 

[Surf Solar]

Damn. :/ Glad to hear that you haven't given up though. I wish you good luck to earn a few bucks with Monsterland. I wish I could provide any help, if it only would be tiles, then that's no problem to make you some. But tiles aren't the only thin what makes a game graphical, heh. (reminds me of this "Tiles Suck" Graffiti I recently disovered in Fallout Tactics ;D ). Anyway, keep it going man!

 

[shihonage]

I feel self-conscious about spamming with contentless "thanks" posts, hence I didn't post until some progress on Monsterland was made.

Your words are appreciated

I locked down Monsterland featureset, and levels are being made faster now. I see it being complete not far from now, and then it's back to Shelter!

After repeated sentiments about how ugly the walls look, I decided to go slightly hybrid route. If I go full-tiles, then engine's cell-based mechanics will become glaring flaws instead of endearing quirks. So a lot of the alphabet soup remains, but some is semi-tiled.

 

[shihonage]

New Monsterland screens:

 

[shihonage]

The shape of things to come.