Out of Pod Experience

I'm curious as to the feasibility of having sprite characters in a game which are, say, 16-bit color, 128x256 resolution, and have 16 horizontal view angles plus all diagonals for a total of 66 angles encompassing at least 32 poses. Anyone with sprite experience able to give me a number crunch on that? Also nice would be a comparison with what sort of detail level on a 3D model would match the sprite in terms of processor and memory costs.

Reaver, my expertise on such topics is at best, extremely limited.

But I will give you something, something I enjoy a lot.

Have this song

Cheers!

Advanced because I've never seen a program use so many angles. Everyone just jumped onto 3D instead.

I assume there are some processing issues with managing a whole lot of image files at once, especially with modern graphics systems being focused on polygons and skins. Just look at Minecraft, for example. It uses 16x16 pixel voxels and even a 5 year old computer has difficulty putting the view distance past 10 chunks (160 voxels radius).

Yes, thanks! That's quite a relief. I was afraid even modern systems would struggle with it.

It might take a lot of work, but it sounds quite doable. I've always liked sprites better than 3D models, especially when everyone first went to 3D. Perhaps someday we will see an indie game using complex sprites like what I described. They would make for a more fluid body appearance of characters, making them appear to have much more detail than they really do. This could mesh well with an old-style appearance in a game, a throwback to the early '90s of gaming. Or it could make for an excellent re-make of Doom 1.

Did not understand a word of what she weas singing, but it's a beautiful song anyways.

Actually no.

It's still 2d. A 3d character can be moved "freely" within the limited degree of freedom the engine works with.
A 2d character can't be rotated freely ... and you can't move it freely.

You can not rotate a sprite character.
One can use tricks to give it depth,
but these work only rather limited.


How do you imagine that would work?
If you rotate a sprite, it's like a sheet of paper.


Tell me your thoughts and forget "3d is better now".
That viewpoint leads nowhere and only limits your own imagination and creativity.

If you see it this way ...
... and assuming we ignore, that ...

... every single viewpoint has to be pre-made ...

... every .... single ... holy ****.


Assuming a full 6DOF that's 360x360x360 different angles.

Okay, let's say we cut that in half ... it's still a huge amount of images.
And we can't cut too much, because of too big steps.

That's just base, please.

And then ... next thing ... for every arm and leg we have to do the same.
Because we need them seperate from the base, for movement!


Then there's the fact that, with your idea, we can't apply lighting.
Because lighting needs depth, but sprites have none.
We would have to bake depthvalues into *every single sprite seperately*.


It's not as easy as you think....


What you seem to think about is having pointclouds flattened as sprites.
That's a whole different story and actually ... works.

*scratches beard* hm...



Nowadays gfx cards can push so many triangles, that we reach
one triangle per pixel easily ... it's totally unsuitable and crap, tbh,
but no one really has any better idea.

Voxels are great, but we're not yet there. (no not minecraft. minecraft is crap)
But I am out of the loop on this one. The math is beyond me, I hate shadercoding,
I don't care about standards and conventions and I rather use assembler for everything,
because coders nowadays are completely clueless when it comes to performance and
"compiler optimises better for speed than a person" ... which can only come from a person
who has no clue about it.

Errr I digress....


And as you mention it ... I have no idea why CCP needs animators to animate the avatars.
It makes no freaking sense to me that they animate characters this way,
as we're not living in the 90s anymore. Motion capturing can be improvised with cheap cameras
and without the need of expensive software ... but oh well.


It's late. I should sleep.
If you prefer Solecist in here, I can get her to post.

Yes, it takes a lot of frames, and probably the only realistic way to produce these sprites in high detail is with 3D models. You could use something similar to motion-capture technology but which turns the images into sprites rather than 3D. It might take a lot more cameras though. Or you could make a very high-detail 3D model and turn it into a sprite, and actually greatly reduce the graphics requirements to render it without losing much detail.

But I like that you mentioned voxels. Perhaps now they are an even better way to get what I want. Instead of having several thousand images per sprite, there could be a voxel sprite. With transparency effects, the edges of the sprite could be properly multisampled, dithered, and have lighting and reflective effects applied even though it isn't rendered in polygons. This would make the character look much more like a person on film and less like a game character.

GGGGRRRRRRRRR

Assuming you would use uncompressed bitmap images you would be looking at 16*128*256 making it 524kb after which you need 66*32 more of similar images so that would make it a 1,1 gigabytes which obviously is quite much and because of it there are a variety of ways to compress images to conserve memory.

Edit: one way to conserve memory in ye olde days was to have a "basic sprite" that was 'edited' on the fly saving tons of memory and is the main reason it took quite a while to have anything related to 3D.