I've just started on a project. It's kinda what I've always wanted to make. Working a little bit at a time, I hope to be done in a year. The biggest inspirations are Knytt, Untitled Story, and the Underside. The fundamental draw is exploring the world, so it must be vast, appealing and alive. Control is also key. Beyond that, I'm aiming for more of a puzzle/item type system instead of a "Metroid powerup" type system. Spud's Quest is a good reference here, although I found that game had its flaws. I don't like the one-screen "scene" format that scrolls when you hit the edge. So I have to break up the large map. If I don't feel really ambitious (which I probably won't), I'm not going to be doing dynamic loading with multi-threading. So the "scene" will be probably about 10x10 screens if I don't break it up some other way. The hook is hopefully the story and interesting characters you meet, as well as exploration. But what is different is I'm going for a high-resolution display, all the inspirations use pixel art. But it is not realism, it's rather something of a painted look. I help with the Secret Maryo project so I have some experience seeing how high-resolution 2D works (except Maryo is vector art). Not many enemies are planned, maybe spikes, lava, swinging spike balls, and turret defenses. Maybe some platforming. Maybe a little bit of maze challenges, because you travel through pipes in the world. I'm using C# and OpenTK. Currently I'm using FBOs, VBOs, and NPOT textures but I may want to switch for older cards/laptops. Ideally, if I had all the time a panda does, I could make it easily mod-able like Knytt Stories and the Underside almost is. But then you have to make tools that others can use and a more flexible code design. However I am definitely using flexible XML formats and possibly lua, but then again I don't really want that level of gameplay complexity. I have to focus on the basics because after that I still have to do art and level design. Hopefully it's good enough I can get some ambient music and sounds, but that is for later. I thought about 2D physics and generalized slopes without tiles, but realized that without those features being the "hook" of the game, I would need more manpower to pull it off.

Is it a game you would play? Does it remind you of any games? Do you think the item/puzzle system would work over the collect/power.up system?

On the technical side, should I use NPOT textures (and shaders?) in the final?

I'm thinking about making the game completely non-violent, but adult-themed (and very little sexual content).

I'm not sure what you need NPO2 textures for. A 2D platformer will be making heavy use of texture atlases anyway; just pack stuff into squares instead of rectangles.

Well if my desired resolution is something like 1280x1024 or 1600x1024 (wide-screen).

I have fairly large tiles planned, so maybe 8 tiles into 960 (=120 pixels). I have somewhat "fuzzy" borders so the actual size might be up to 140 pixels (not sure if this is going to work). In such a case, assuming I want a texture as big as or larger then rendered resolution, I guess I could pack 14x14 tiles on a 2048px texture.

The main character's frames would not be uniform, so maybe I could fit 10x8 frames on a same size texture.

So, I guess you are right.

I have a current "small tree" that is 419x676px, but a large tree object might be half the screen (i.e. 640x512px). With some careful planning I might be able to pack some bushes and fences in a 2048 texture.

You won! Thanks.

I was thinking big tiles just so it would be easier to design an immense world. Knytt Stories is 10x25 tiles so 8x10(12) isn't too far-fetched.

That mixture of small tilemaps sounds very interesting. Another idea is to have a completely free map system and just have a "tile" mode for the editor to make it easier to design. This is where Maryo is headed.

I'm not going to be worrying about optimizing too much right now because it prevents me from finishing projects. Like I subdivided a tilemap into smaller regions as display lists for rendering, which is probably overkill. And I realized it's not so important as the other game I help with has a separate texture for each tile and doesn't batch at all and it performs decently.

There will not be sexual content but there might be some nudity. The adult themes are more sophisticated.

I'm going to go for a combination stencil tiles, texturing, and separate outline.

Thought about arbitrary placement, but stuck with my idea that tiles are still suited unless you really have reason and a small enough environment to carefully craft each scene.

I'm probably not going to finish the project, just don't have the interest required to carry through code, art, and design. The scope would have to be scaled down a lot. Not to mention online OpenGL documentation is a bitch.

I might write a book instead.

I don't call myself an "artist" but here is the art for the project if you are interested.