[Discussion] What are you working on?

I'm currently working on a typical Pokemon fangame. I currently do not have much progress on the maps and scripting, although I do have over 50 new Mega Evolutions and three new types. The only problem is that I am just a beginner, and I'm not incredibly skilled in spriting, mapping, or scripting. I am good with ideas, if that matters, though. It would be much easier if I actually had skilled people to work with.

I'm working on a game with an original story called Uniques United. Right now, I'm balancing out the magic system, as I've recently implemented a "Draw" ability called Scribe, like the Draw from Final Fantasy 8. This game uses original sound design, I made every sound effect and ambient track myself.

Working on pokmon sky browser:

So far having issues with little things , but when done I will re-release the source !

Original Source :
http://forum.ragezone.com/f705/release-poka-mon-sky-browsergame-1039028/

Working on the Town Map for Gensokyo in Faith & Prayer Version. Parts of it are slated to be cut out and placed on a separate map later on, right now they're just there because I didn't get around to removing them yet.

I'm working on adding the Gen VI Pokémon, as well as new TMs and new moves. Scripting isn't my forte, so I intend to save doing that for as long as possible.

Took a short break to enjoy the christmas break and work on Nax- which has been going really well, lots of backend code changes and such.

I'm still very undecided on the storyline, I want it to somewhat take inspiration from One Piece, but instead of travelling the world, you travel the universe. trying to reach the mythical "center of the universe" (which if you pay attention to astrophysics, doesn't actually exist). You're gonna be travelling planet to planet, where there is usually a tyranical leader, or crisis going on which you will help with.

Also resumed Super poekmon eevee edition development, 0.75 is gonna be out in a month or so (probably late march?). I'm hoping it goes smoother than 0.74 due to the maps already being completed. And no big systems taking up my time. I'm really excited about the summoning system, which is a huge throwback to Final Fantasy. Think the FF summoning system, but with legendary Pokemon.
I'm hoping i finish it this year, but I've been saying the same thing for the last 2 years lol.

Writing a shader for Unity which constrains an object to a fixed color palette and can allow normal maps. This could be used really well to recreate the butt-load of shaders that Pokemon XY/ORAS use to make their models look sprite-like, or for more extreme effects. When it's done, it'll be free to use with credit.

Java's Missingno. said:

Writing a shader for Unity which constrains an object to a fixed color palette and can allow normal maps. This could be used really well to recreate the butt-load of shaders that Pokemon XY/ORAS use to make their models look sprite-like, or for more extreme effects. When it's done, it'll be free to use with credit.

Click to expand...

Oh wow, that sounds really neat. Can't wait to see how it turns out!

Ashedragon said:

Oh wow, that sounds really neat. Can't wait to see how it turns out!

Click to expand...

Well, here's a look!

Used a palette roughly corresponding towhat would be found in the RBY generation(4 colors, high contrast, and some dithering), applied to a model of Omanyte direclty ripped from Pokemon X/Y. I also have one which takes normal maps(which aren't present on the XY/ORAS), so that you can get images to 'pop' even more. You can also tell that unlike the official shaders, this one doesn't produce an outline. Part of that is because I don't like the way the official shaders calculate an outline(they calculate it based on a difference in color, so sometimes outlines will randomly appear on jagged parts of a model and then vanish in the next frame. Instead, I'm hoping to simply create a cel-shading outline outside the model which is geometry-based. The easiest way to do this is to simply make a slightly larger model and flip the normals inside-out(for instance, this is how Rotom gets that aura effect) and set it to all black. Since those normals face away from the camera, you can see "into" the outline model to the underlying model, so it looks like there's an outline outside of the object. This runs into other complications with more complicated geometry. I think Omanyte's eyes would have some difficulty with the geometry-based outline, but I don't know for sure. (Blast it, Jim, I'm a designer, not a 3d modeler!)

BadSamaritan said:

I beg to differ ;)

http://prntscr.com/64bq9g

In fairness, it seems not all pokemon have them. Also some pokemon like Wailmer obviously have a spec map in addition to the normal to give it that scaly look.

Also I don't think the outlining works like you think(though I'm not 100% sure). Not only would that way push the 3ds beyond what are already incredibly high poly models, it's far simpler to do what 3ds max(which nintendo used for ds, 3ds, wii) and blender does, which is simply to generate curves and "stroke" them. Because unlike the way it works with Rotom, you don't always want an outline on every single visible seam. For instance if you're viewing a model from the side, you might want the outline of an arm to taper off as it connects to the body.

But ANYWAYS, the shader you showed is incredibly impressive, and it seems like you know what you're doing. I can't wait to see the finished product!

Click to expand...

Well, if you *beg* to differ ;P Very cool, though - I hadn't seen spec maps nor normal maps on the (admittedly few) rips I've poked around with. as for the outline shader, from what I'm aware, it has a predefined 'palette' of colors(for instance on Feraligatr, you've got some reds, blues, tans, and black/white), and the textures are mapped to that palette. afterward, on a per-pixel basis, the shader traces the outline of the Pokemon overall(unless it's got inverted normals like Rotom). Following that, it steps through the pixels and marks off outlines between 'sharp' changes in pixel color(you can see this with spiky Pokemon like Feraligatr, gyarados, magikarp, and tyranitar). Since this is done on a per-pixel basis, it would lead to the lag you sometimes observe in XYORAS, such as battles with Pokemon with other geometry effects in play(for instance, Moltres has particle effects on its wings, Klefki has rigidbody physics on the keys, Goodra actually uses metaballs for dripping).

With inverted normals, you don't get those 'internal' outlines quite as cleanly, but instead would get a much more cheaply-rendered outline, which is great for performance if you're willing to sacrifice a bit of finesse in your art assets.

BadSamaritan said:

Ah, very cool. Curiously enough, how do you know so much about how the game does this stuff? I mean, because I'm a 3D artist I can guess many of the ways it does something, but I definitely have no idea how it's technically done.

Also yeah, the normal/spec maps seem packed in with the diffuse maps(presumably because they are just placeholders for future games/movies that aren't cell shaded and will use the maps more), so they don't appear in regular file viewers. It's only through PaintTool Sai that I actually find them.

Click to expand...

Mostly it's through lots of observation with a relatively trained eye - I'm a 3rd year BSCSGD at DigiPen, so I've learned a little bit of everything(art, psychology, team dynamics, and a whole lot of game design and computer science), though I am by no means an expert on anything! :P I've also been writing some shader code from time to time, and thinking about ways that I'd recreate pixel art dynamically. Usually the stuff I know about is stuff I've tried to implement myself at some point or another - dynamically shaded sprites, terrain generation, particle effects, that sort of thing.

Interesting; I would guess it's a hardware issue with the 3DS, actually, and that the art team made all the textures that could possibly be used for a model before the shaders were written. Then, since the shader wound up being expensive computationally, probably left the art assets in-game because they didn't cost much in terms of game space. (We actually still have some REALLY old art of a dragon in my class game project that we were using to demonstrate that we had animated sprites back at engine proof)

I've always been annoyed at how bad the AI is in Pokemon games, and while it was improved in the newer versions of essentials it still didn't satisfy me. I've been working on coding a better AI in the last couple of weeks and while there is still a lot of work to do, I am happy with the progress I have made. I actually uploaded a youtube video of a test battle I had with it, but I can't link it here because I don't have enough posts to post the link.

greenOak said:

I've always been annoyed at how bad the AI is in Pokemon games, and while it was improved in the newer versions of essentials it still didn't satisfy me. I've been working on coding a better AI in the last couple of weeks and while there is still a lot of work to do, I am happy with the progress I have made. I actually uploaded a youtube video of a test battle I had with it, but I can't link it here because I don't have enough posts to post the link.

Click to expand...

I'd really be interested in seeing how you implemented it - does it conform to a specific strategy or use something more akin to fuzzy logic?

Java's Missingno. said:

I'd really be interested in seeing how you implemented it - does it conform to a specific strategy or use something more akin to fuzzy logic?

Click to expand...

It's more or less the same way it is implemented currently in essentials. Each move gets assigned a score - which the way I set it up is directly related to the damage output or equivalent in the case of a move like toxic, thunder wave, etc. Each switch to another pokemon is also assigned a score depending on the damage (or equivalent) that will be taken on switching in as well as the damage that will be done to each pokemon on future turns. Laying down entry hazards like spikes and stealth rock are assigned a static score. I've written separate code for recovery moves and stat uping moves which again assign a score. My AI then simply chooses the move or switch with the highest score.

I did use one pretty big shortcut though in that the AI knows the users team. I didn't feel like trying to write the code where the AI has to "guess" the users team. I also felt this wasn't too big a problem as the user can also know the AI teams in pokemon games.

greenOak said:

It's more or less the same way it is implemented currently in essentials. Each move gets assigned a score - which the way I set it up is directly related to the damage output or equivalent in the case of a move like toxic, thunder wave, etc. Each switch to another pokemon is also assigned a score depending on the damage (or equivalent) that will be taken on switching in as well as the damage that will be done to each pokemon on future turns. Laying down entry hazards like spikes and stealth rock are assigned a static score. I've written separate code for recovery moves and stat uping moves which again assign a score. My AI then simply chooses the move or switch with the highest score.

I did use one pretty big shortcut though in that the AI knows the users team. I didn't feel like trying to write the code where the AI has to "guess" the users team. I also felt this wasn't too big a problem as the user can also know the AI teams in pokemon games.

Click to expand...

Ah, I see. The only difference between what you described and what I was envisioning is the fact that you take the highest-scoring option. For Beat Down, I actually implemented a weighted score system for the AI, which weights the different actions available(attack, retreat, advance, jump, etc.), and then randomly selects, with each action's probability weighted by its score(so if jump has a score of 10 whereas attack has a score of 5, it's twice as likely to pick jump), which makes the AI overall less predictable but still have some semblance of strategy.

Java's Missingno. said:

Ah, I see. The only difference between what you described and what I was envisioning is the fact that you take the highest-scoring option. For Beat Down, I actually implemented a weighted score system for the AI, which weights the different actions available(attack, retreat, advance, jump, etc.), and then randomly selects, with each action's probability weighted by its score(so if jump has a score of 10 whereas attack has a score of 5, it's twice as likely to pick jump), which makes the AI overall less predictable but still have some semblance of strategy.

Click to expand...

Yeah I want the AI to use some mixed strategies too, but that is farther down the line for me.

I don't see why the game wouldn't be using those normal maps already - look at how Sneasel's right ear has very slight shading differences compared to its' head.

As for the bright white/black maps, the ones corresponding to eye textures appear to be used as an opacity map for the eyes where a Pokemon has their irises floating on top of their face. Where they exist for the body I don't know but I have a feeling they're implemented as a kind of shadow bias map, even though most Pokemon have shadows baked into their diffuse textures there are things like the band of shading on Greninja's arms which look like they match those maps instead.

You'd get a definitive answer though if someone with the right tools could dummy out one of the maps with a blank/solid color image and compare the results.

COOLTRAINER♂ said:

I don't see why the game wouldn't be using those normal maps already - look at how Sneasel's right ear has very slight shading differences compared to its' head.

As for the bright white/black maps, the ones corresponding to eye textures appear to be used as an opacity map for the eyes where a Pokemon has their irises floating on top of their face. Where they exist for the body I don't know but I have a feeling they're implemented as a kind of shadow bias map, even though most Pokemon have shadows baked into their diffuse textures there are things like the band of shading on Greninja's arms which look like they match those maps instead.

You'd get a definitive answer though if someone with the right tools could dummy out one of the maps with a blank/solid color image and compare the results.

Click to expand...

Main reason to skip using them would be hardware - normal calculation for dynamic lighting is fairly expensive, and the game has huge framerate drops as-is in certain circumstances. That said, I don't know for sure one way or the other, and my thoughts on how it's done -are- based on casual observation and could very well be completely off. :P

Somebody should make a randomizable AI script. Like their strategy changes if you battle them again or something.. That WOULD BE AWESOME.

schoooooool!

Ugh, it has not been as entertaining as I had hoped. :(