I think I will go for outcrop corner. Outgrowth just sounds wrong. (^^)
Yeah for sure. I do like to hear about these obscure/domain specific terms though.
Last edited:
-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
Indie Game Development |OT 2018| Come, help, learn, show and tell!
- Thread starter Popstar
- Start date
- Status
I'm starting to get my auto tiling into a robust place. Figuring this stuff out is a real pain in the ass but also rewarding and fun.
The way this works is that I have a density grid(basically a "3d" array filled with zeroes and ones) - which I freely can paint into - that I pull marching cubes from to get a final tile. Unfortunately the marching cube approach only gets you so far(many related but distinct tiles/shapes end up having the same encoding). To get the correct tile I also need to look at the cube above as well as figure out what kind of piece(is it the full, low, mid, or top piece?) either tile should be. One useful aspect of this approach is that in the case a hole appears in my model all I have to do is to look at the bitfield to figure out what sort of piece I am missing.
With tiles ready the next step is to to stitch them together(technically the normals). Getting this right, while also doing it efficiently(doing it offline is trivial) is a major headache that affects everything from the original 3d model, to asset processing and in engine stitching. Considering all that trouble, It makes a lot of sense why so many 3d-tile games go for the flat shaded look. It's nice to have it working now though. Hopefully no bugs!
now I just hope that I won't start to want diagonal cliffs..
Last edited:
I've been working on the character lighting/story elements for Thieves in the Night:
The lighting uses traditional cel-shading with some diffuse only fill lights for additional depth. At the start of the game, the player is engaged in low-level crime with crude weapons. It'll progress to the more sci-fi looking stuff shown in some of the earlier demos.
My wife and kids helped me out a bunch with look and style. My older bro drew the title.
The lighting uses traditional cel-shading with some diffuse only fill lights for additional depth. At the start of the game, the player is engaged in low-level crime with crude weapons. It'll progress to the more sci-fi looking stuff shown in some of the earlier demos.
My wife and kids helped me out a bunch with look and style. My older bro drew the title.
Last edited:
Maybe I'm misunderstanding what you're doing, but are you using quads rather than tris? Moving to rendering Tris instead of Quads first, and then 'smoothing' would probably help with auto-solving diagonals?
This is all in engine where everything is a triangle so I'm not sure what you mean. The problem with adding diagonal cliff tiles is that they mean a lot of extra work and complexity.
Last edited:
I used Ren'py for mine and was fairly happy with it. You can also get some pretty cheap asset packs for unity that will give you a bit more control. Feel free to pm me if you need help or tips.
I'm starting to get my auto tiling into a robust place. Figuring this stuff out is a real pain in the ass but also rewarding and fun.
The way this works is that I have a density grid(basically a "3d" array filled with zeroes and ones) - which I freely can paint into - that I pull marching cubes from to get a final tile. Unfortunately the marching cube approach only gets you so far(many related but distinct tiles/shapes end up having the same encoding). To get the correct tile I also need to look at the cube above as well as figure out what kind of piece(is it the full, low, mid, or top piece?) either tile should be. One useful aspect of this approach is that in the case a hole appears in my model all I have to do is to look at the bitfield to figure out what sort of piece I am missing.
With tiles ready the next step is to to stitch them together(technically the normals). Getting this right, while also doing it efficiently(doing it offline is trivial) is a major headache that affects everything from the original 3d model, to asset processing and in engine stitching. Considering all that trouble, It makes a lot of sense why so many 3d-tile games go for the flat shaded look. It's nice to have it working now though. Hopefully no bugs!
now I just hope that I won't start to want diagonal cliffs..
That's cool. All the tiles are generated/modeled right? Why do you need to stitch them together?
Because otherwise they end up looking like this:
This is pretty good for having no normals blending. With less uniform models the artifacts are worse and show up in more places. These artifacts happen because tiles often have different normals on their borders, and if the normals aren't the same you'll get a discontinuity in the lighting(this is how smoothing groups work).
It's interesting to me that almost every game I've come across that use these kind of 3d-tiles also use flat shading. They usually run on low-end platforms so maybe it's just too much trouble making the stitching work in realtime. Also with flat shading you don't have to worry about how the geo will f-up your normals, which makes the tiles a lot easier to model.
Last edited:
That's very cool. Frankly if the map is relatively straightforward like that specific one, I think you could get away without zoom or rotation (I'd also imagine rotation would either be limited or need an animation for "changing where I'm holding the map from").
Thanks! Yeah you're right, I definitely want to keep the rotation / move very limited. It's actually less for functionality but more for that sense of 'tangibility'. At the moment it feels more like a piece of UI even though it's in world. Being able to slightly twist it with the thumbsticks would be a nice bit of polish I think, beds it more in the world as something the player is holding.
Zoom I think might be more essential when the world grows but you're right, right now with this level of detail it's not a huge issue :)
That's the easiest to read car map I've ever seen >:P
The mood is so strong here. Very lovely.
Noice
Also, I know someone was in here asking about what programs to use for pixel art the other day -- I had a crash this week and had to reboot my computer, and saw that the program I use, GraphicsGale, is free. It happened about a year ago, apparently, but hey, better late than never. Definitely the best sprite art program I've worked with.
Because otherwise they end up looking like this:
This is pretty good for having no normals blending. With less uniform models the artifacts are worse and show up in more places. These artifacts happen because tiles often have different normals on their borders, and if the normals aren't the same you'll get a discontinuity in the lighting(this is how smoothing groups work).
It's interesting to me that almost every game I've come across that use these kind of 3d-tiles also use flat shading. They usually run on low-end platforms so maybe it's just too much trouble making the stitching work in realtime. Also with flat shading you don't have to worry about how the geo will f-up your normals, which makes the tiles a lot easier to model.
What's stopping you from having the normals along the tile borders set correctly in advance though? The "outcrop corner" and the piece above it would have incorrect looking normals on their own but together would always look smooth.
I thought you were procedurally generating meshes from an array of vertices, which is usually done by creating triangles, then turning those into quads for 'tiles' - in which case you would be able to do 'diagonals' by checking if a triangle was valid (which would presumably only be valid on certain axes), but if you're not doing that, thats not much help.
Well almost every tile can fit with many different tiles so there is no one correct way to blend the normals, and using the average of all combinations would get rid of the discontinuity but still create bumps. Consider an inside or outside corner, it can fit with both flat walls or another inside or outside corner. But the initial normals on the border point in different directions for all these pieces.
Maybe you mean that I could store all combinations in a table?
Having a small breather between work, job applications, and helping my wife with projects, I was able to make a little time for my next game I want to make. Was able to get placeholder art and systems setup so I can control the environment and NPCs from one object. Was able to make some blueprints that auto fill designated spaces with dancers ^_^
The end goal is to have it all tie into whatever music is playing at the moment. Enjoy horrible clip art in a cube club:
The end goal is to have it all tie into whatever music is playing at the moment. Enjoy horrible clip art in a cube club:
Feel free to let me know if I'm not being helpful since I don't want to waste your time, but each color is a different tile right?
Every tile except the green and the one above the green will always have seamless normals regardless of how they are placed. If your game didn't support the green outcrop corner then you wouldn't need to mess with stitching. Your example picture without stitching shows this.
To fix this, the normals along the top edge of the green piece could be precalculated to fit with the wall above them. Then the same would be done to that wall and its bottom edge. You might need to add another tile to your setup to make it work but then you don't need to do any stitching.
your posts are helpful, if nothing else it forces me to think through assumptions I've made and in best case I realize I'm doing something stupid/unnecessary.
Here I use a different tileset to better illustrates the issue
Col #1: no averaged normals
Col #2: normals averaged with neighbor
Col #3: normals averaged with the neighbor from the row above
Col #3: normals averaged with the neighbor from the row below
As you can see, normals are only correct in #2 where they are averaged with the tile actually sitting next to it. That's what you'd expect unless the normals are parallel on both sides of the edge.
That picture I posted does a bad job of showing the incorrect normals, and that's mostly because the normals are or almost are parallel across borders in that particular tile set. Enforcing that as a rule would put heavy restrictions on what the geometry has to look like and I don't plan to do that.
Here I use a different tileset to better illustrates the issue
Col #1: no averaged normals
Col #2: normals averaged with neighbor
Col #3: normals averaged with the neighbor from the row above
Col #3: normals averaged with the neighbor from the row below
As you can see, normals are only correct in #2 where they are averaged with the tile actually sitting next to it. That's what you'd expect unless the normals are parallel on both sides of the edge.
That picture I posted does a bad job of showing the incorrect normals, and that's mostly because the normals are or almost are parallel across borders in that particular tile set. Enforcing that as a rule would put heavy restrictions on what the geometry has to look like and I don't plan to do that.
Last edited:
your posts are helpful, if nothing else it forces me to think through assumptions I've made and in best case I realize I'm doing something stupid/unnecessary.
Here I use a different tileset to better illustrates the issue
Col #1: no averaged normals
Col #2: normals averaged with neighbor
Col #3: normals averaged with the neighbor from the row above
Col #3: normals averaged with the neighbor from the row below
As you can see, normals are only correct in #2 where they are averaged with the tile actually sitting next to it. That's what you'd expect unless the normals are parallel on both sides of the edge.
That picture I posted does a bad job of showing the incorrect normals, and that's mostly because the normals are or almost are parallel across borders in that particular tile set. Enforcing that as a rule would put heavy restrictions on what the geometry has to look like and I don't plan to do that.
I've been assuming that you'd ideally want parallel normals. It'd be the most mathematically accurate method and would work with all geometry you've shown. In this set the normals along the edge would just need to have no left/right component. If you plan to do more complicated shapes then maybe you could do a hybrid approach. Mark certain tiles as "pre approved neighbors" that fit smoothly together without stitching and then only use stitching on the few tiles that need it.
I see where you are coming from now. The production set for say a rocky cliff has more irregular shapes on the slope.
I guess I'm not sure I understand how parallel edges would work for the tiles in the last picture I showed. With the corner piece the faces next to to the edge are not perfectly perpendicular to the edge, so the normals are not parallel to the edge. If I pre-process them to align they will be incorrect.
Yeah I already do this to some degree, I simply skip the averaging stage when I know the normals will line up. Also the stitching is implemented using precomputed, sorted index tables for each border/tile combination. Finding the vertices is constant time and then it's just a few wide instructions per vertex pair to average the normals and tangents so overhead is minimal.
Last edited:
If you also want rocky cliffs then that sure would make it harder.
The normals are not parallel to the edge because your models are low poly. Pre-processing them to be parallel would be incorrect from a mesh point of view but would be accurate to the topology you're representing. Might make the scene look slightly better even (or slightly worse since realism isn't the end goal).
That's smart.
Work on Project Moon Magic has mostly been backend stuff right now, mainly refactoring the code so it's not so much of a spectacular mess, and switching over to Unity 2018.2 to see if it broke anything, which, fortunately, did not happen. I also got an asset that generates text-to-speech audio files for prototyping dialogue stuff. Speaking of dialogue, I've also added a placeholder indicator for dialogue responses, and I've implemented a system for conversations that allows the player to return to a previous 'main' conversation after it is interrupted by a 'side' conversation (such as a scene-based dialogue).
I also got a email back from Team 17 (finally) but sadly the same result as Devolver (positive feedback, but rejected due to poor lineup fit).
Though I guess I've got something - bodies. Basically, the current prototype starting area is supposed to be filled with corpses of personnel and security bots, so I bought a cheap placeholder humanoid model, modified some existing animations I had on hand to make static corpses, and, well, the starting scene now actually sorta looks like the site of a massacre.
I also got a email back from Team 17 (finally) but sadly the same result as Devolver (positive feedback, but rejected due to poor lineup fit).
Though I guess I've got something - bodies. Basically, the current prototype starting area is supposed to be filled with corpses of personnel and security bots, so I bought a cheap placeholder humanoid model, modified some existing animations I had on hand to make static corpses, and, well, the starting scene now actually sorta looks like the site of a massacre.
Last edited:
Been working on this enemy, alternating with less visual stuff. I think I should have tweeted it at 3x pixel size rather than 4x, but eh.
So I listened to you guys and changed my game name! Better now than later. Thank you guys! I obviously had to update the trailer and added a lot of new and updated scenes!
Looking absolutely, positively insane; can't wait to play it. I may have asked this before (and if not I'm sure you've been asked by others), but are you planning on having a Switch release?
Also, from your earlier trailers I didn't realize mechas played a central part in the game (I thought they were kind of a one-time thing, not a core game mechanic). This is kind of the JRPG I've been dreaming of since the 90's. :D
Thanks, yeah the mechs are a core mechanic! :) Switch will be a stretch goal in my campaign. Everyone really wants to play the game on it.
Wow, this looks really incredible! Did you make this all by yourself? O.O
Quick question as a nondev, is concept art or promotional paintings allowed on Steam store pages?
I've noticed a lot of games don't have this, and I think it would be a really awesome way to make your game stand out.
Wondering if offering a promotional illustration for soundtracks/PDF manuals to dev teams would be a good way draw in an audience that other wouldn't check out your game. Kind of like the covers used for Final Fantasy or Castlevania.
I've noticed a lot of games don't have this, and I think it would be a really awesome way to make your game stand out.
Wondering if offering a promotional illustration for soundtracks/PDF manuals to dev teams would be a good way draw in an audience that other wouldn't check out your game. Kind of like the covers used for Final Fantasy or Castlevania.
Thank you guys! I really appreciate it!
Thx! Yeah, except for the music which was done by the game's composer Eddie Marianukroh.
I guess most of us barely have the budget to make the game itself, hahah. I mean, some promotional art seems kind of necessary, but I can't draw for shit, and I can't pay anyone to do it, so... Same thing happens with music, which is why I use public domain tracks.
I believe that Valve put down a "Game Screenshots only" policy a few years ago after someone rocked the boat a little too hard by using only Pre-rendered out-of-engine Bullshots as their screenshots and it went viral by way of Jim Sterling.
Interesting, I didn't know about that. I guess there's nothing preventing you from circumventing it by using that art as intro / title screen / cutscenes, and it's probably the most efficient use of art anyway.
I've lost a bit of focus on my project after switching to a FPS. Just don't know what I wanted to do with it. Probably going to focus less on the hook but have it be a main ability.....then I realized I had a marble game and could totally just do some Metroid Prime shit, so I think I will go that route.
The package itself is "2D Pixel Perfect" (no "camera"). Have you installed the package manager UI? I remember having to install that before I could install the pixel perfect camera. That said, I ended up not using the pixel perfect camera itself, for several reasons.
Why not use a grey scale heightmap? Might be easier to design models with. Unless you want the holes inside the model?
I'm starting to get my auto tiling into a robust place. Figuring this stuff out is a real pain in the ass but also rewarding and fun.
The way this works is that I have a density grid(basically a "3d" array filled with zeroes and ones) - which I freely can paint into - that I pull marching cubes from to get a final tile. Unfortunately the marching cube approach only gets you so far(many related but distinct tiles/shapes end up having the same encoding). To get the correct tile I also need to look at the cube above as well as figure out what kind of piece(is it the full, low, mid, or top piece?) either tile should be. One useful aspect of this approach is that in the case a hole appears in my model all I have to do is to look at the bitfield to figure out what sort of piece I am missing.
With tiles ready the next step is to to stitch them together(technically the normals). Getting this right, while also doing it efficiently(doing it offline is trivial) is a major headache that affects everything from the original 3d model, to asset processing and in engine stitching. Considering all that trouble, It makes a lot of sense why so many 3d-tile games go for the flat shaded look. It's nice to have it working now though. Hopefully no bugs!
now I just hope that I won't start to want diagonal cliffs..
It's such a weird feeling toiling away at a game for months or years, and then finally getting your first AI working. Suddenly this solitary experience feels alive. Even if the AI is super early or completely broken, there's something special about having another living creature in your world.
I haven't installed any package manager UI, but I do have a package manager window. I've watched some videos of others using the 2D Pixel Perfect and am looking for the correct package, it just doesn't seem to be there.
That being said, do you have any suggestions on what to do to avoid the pixel shape shifting stuff since you didn't use the camera yourself?
I coded my own solution, it took a bit and I may have not done it if the pixel perfect camera existed when I tried it, but now I actually prefer it. It locks the resolution / camera zoom at a perfect multiple of the game's pixel size, so there are no distorted pixels at any resolution, plus you get "smooth" scrolling (rather than pixel-by-pixel).
My specific solution means that the camera can show more or less stuff depending on the resolution, though (specifically, it tends to show a slightly larger area when the resolution is not a perfect multiple of 480x270). For some games this may be a dealbreaker, in mine it's not particularly problematic (alternatively, you could simply drop some black bars).
I can give you the scripts if you want them, let me upload them to my Google Drive.
Been working on Everything For Momo since the start of this month. The game is being developed in RPG Maker at the moment although once a rough prototype is up and running I might want to move the whole thing to Unity so we can port the game to consoles as well. We've got three really talented pixel artists working on this and we're finalizing our weather effects and characters.
Since this is my first game, I'm trying to keep it somewhat simple starting with the limited color palette. This is going to be a straight narrative adventure game similar to To The Moon. If I actually manage to finish this one day, I think I will have gained enough experience to do something a little more complicated.
Since this is my first game, I'm trying to keep it somewhat simple starting with the limited color palette. This is going to be a straight narrative adventure game similar to To The Moon. If I actually manage to finish this one day, I think I will have gained enough experience to do something a little more complicated.
Last edited:
Here you go:
https://drive.google.com/open?id=1ZpZgp8PyaTX21bZJ1PYY5Luf1v8mPhxB
I think I ripped out all references to my other classes (and hopefully didn't break much in the process), but I've not tested this version. If it doesn't work and you still want it, let me know and tomorrow I'll try to set up a minimal test project and do it a bit more properly.
I think I need to buy a new laptop (again), but I haven't lifted a dev finger since I got my new game dev job. I really want to but I'm exhausted from the new city, job, and overtime already. When things have settled though, I'll be back in full force :) First order of business is probably to complete as many online courses and also try to finish my previous project.
- Status