General > General Discussion
Ramblethread! A brainstorm approaches!
Arachne:
--- Quote from: Ai on March 06, 2009, 10:40:35 am ---It also mentions something else which is important for newbies, IMO: because of gamma, the midpoint of the RGB values of two colors usually *ISN'T* a 50% mix of the two. (eg black #000000, white #ffffff -> wrong result = #7f7f7f, right result = #bababa). Knowing this can help avoid ineffectual shading as well as ugly AA.
--- End quote ---
This is a good tip. And it also makes decent monitor calibration a must.
I think what's mentioned in this post by Ilkke, with the principle of looking at diagonals, is a very good way to check if your AA makes sense. Most of the time, though, I have to compare what I'm working on with the previous step to make sure I don't over-AA. If it's actually blurrier than what I had, I have to redo it, and I haven't found a better way to make the AA look good than to go through this tedious process. :blind:
Ai:
--- Quote from: Arachne on March 06, 2009, 11:32:06 am ---
--- Quote from: Ai on March 06, 2009, 10:40:35 am ---It also mentions something else which is important for newbies, IMO: because of gamma, the midpoint of the RGB values of two colors usually *ISN'T* a 50% mix of the two. (eg black #000000, white #ffffff -> wrong result = #7f7f7f, right result = #bababa). Knowing this can help avoid ineffectual shading as well as ugly AA.
--- End quote ---
This is a good tip. And it also makes decent monitor calibration a must.
I think what's mentioned in this post by Ilkke, with the principle of looking at diagonals, is a very good way to check if your AA makes sense. Most of the time, though, I have to compare what I'm working on with the previous step to make sure I don't over-AA. If it's actually blurrier than what I had, I have to redo it, and I haven't found a better way to make the AA look good than to go through this tedious process. :blind:
--- End quote ---
Well honestly I think it really is all maths, especially when you're focusing on exact rendering as you seem to suggest above; you're basically just emulating an gamma-correct optimized AAed-polygon rasterizer (note that most polygon rasterizers are substantially wrong because they aren't gamma correct - Don't base your idea of correct AA on these!.)
Anyway, perhaps I should have also mentioned how I arrived at the 'correct' value: I gamma-adjusted the RGB values to gamma 1.0 (linear) before blending them, and gamma-adjusted them back to standard RGB gamma.
Roughly:
--- Quote ---0.0 #black
1.0 #white
blackcorrect = 0.0 ^ (1/2.2)
whitecorrect = 1.0 ^ (1/2.2)
result = (black + white) / 2
finalresult = result ^ 2.2
--- End quote ---
The above is inexact, and close enough to perfect for pixel art purposes. I don't actually calculate, I eyeball it (it's good enough though that I hardly ever need to redo it); But this is because I have used examples of polygons rendered with accurate gamma treatment as examples to develop my sense for this; IMO pixel art itself is not a useful training aid here, nor anything but a plain flat gamma-correct antialiased shape rendered by computer. (Blender is a nice tool here, with a shadeless white material on a black background it's easy to experiment and get accurate renders.)
Helm:
Doppleganger: I do think something like this needs to be done too. I don't know if I can do it but I think Pixelation on the whole can do a good enough job of it. To answer your questions:
I will explain perfect lines more.
As to the abstraction and who my target audience is: I've noticed most pixel art tutorials online are basically 'look how I am doing pixel art, try to do the same'. I think this avenue of approach is best left to video tutorials (now that the technology to do this has caught up with us) and that a 'real' tutorial should best approach the fundamentals of the artform from the abstract towards a workable hypothesis. Ostensibly, this has with one leg on academia and one leg on practical application and I will not say this might not be a bit troubling (I can hear the 'lol pretentus artfagllool's from now) for some but this is my target audience: people that have tried pixel art 'blind' for a few months/years and are sensing a lot of the truths of pixel art but canno formulate them exactly. These people that want a theoretical framework that will help their work process and augment their already existing artistic instincts. That's the target audience. Not the "10 easy steps to pixel a street fighter sprite" people, whom are best serviced by numerous other tutorials on the internet already.
I do not intend to cover every aspect of doing pixel art because a lot of that is just doing art art. I want to cover the specific things relevant to pixels in particular that you would not get from reading a good book on art theory for example. I will not cover for example, color selection or anatomy or even rendering in the abstract. I will cover pixel clusters, banding, dithering and the partaining effects of these things. Along with anything else you guys will find worth saying on the side. In short, this will not be all-encompassing. I can't write a book on this, I am not equipped enough. And I can't make you guys write a book on this all together either, heh. But these are thoughts I've been trying to pull all together for quite a while now so this time is good as any. I want to write this and then when someone bands I want to link them to this so they can get a full understanding of what I am saying.
Gil - and other people - : As I said - and perhaps I should stress this more later on in the text - it's a juggling act between expressing 'infinite res' and also conveying what you're trying to render. If you are drawing something with a hard edge that theoretically needs a specular that would be less than a pixel big but you don't have the colors to fake subpixelling then tough shit, singlepixel specular it is. Note in fact how in your example, the shinier the specular, the BIGGER the pixel in question looks around the various other more elegant forms. In fact I postulate that a two-pixel cluster of a less bright color will appear slimmer than that single superbright fatpixel. I am not in a computer with pixel art apps so you'll have to test this out for me. I will write more on the subject when I'm at home. Your further observations are very welcome.
Arachne: curves can never be perfect and in my opinion between perfect lines and more calculated curves the pixel artist should go with the perfect line and then attempt to convey the added curvature of the object he's trying to render via AA, if available. But when we get to that proper you should write about it more.
About looking at the pixel you placed and making a judgment call if it made the piece better than before, that sort of thing goes without saying and really I can't base such a text on these methods. It would just deteriorate to "place pixels until you like what you did" well... I'm sure most artists have tried this method extensively. Hopefully through codification why one likes this placement of pixels more than that one will be revealed. A lot of your trouble with aa-ing and then not liking it and redoing it comes from when you have a lot of colors, a small space to antialias and you also want to convey a sharp shape there so the aa-ing you do touches the contour of your cluster and hugs it so much it begins to band. Then you look at it at 1x zoom and it looks blurry, so you remove the much aa. My further text will posit that good clusters of pixels don't need a lot of aa to begin with and that the artist should first learn extensively how to aa with a single shade between two colors with single pixels or clusters of two pixels (we all know what I mean, here) before they go into tapering and blending their shapes with more shades. The problems begin to occur, I mean, when the artist is trying to combine their blending of clusters with their aa-shaping of clusters. More on this when I am on my own computer.
Thank you for the feedback so far. More soon.
skw:
Very interesting topic. I'm sure I'll learn a lot from it!
Gil: I found out that with such an approach to the detail, pixel art is all about shortcuts and signals rather than presenting the form as it is in the real world. That means, an artist symbolizes the shape of a given form by means of subtle details like e.g. the 1-pixel highlight you have going in your example. I'll pull it forth, tweaking colors a bit, and leaving out the second highlight (colors have a heavy impact on how the piece comes out, and how well the tricks work):
Having three available colors for the faces, it's a waste not to use the brightest one at one of the faces! :)
Helm, I remember you've already covered banding in a very interesting academic fashion at some point back in time -- I can't, however, find the right post. I recall it was some Santa Claus piece.
blumunkee:
--- Quote from: Helm ---...As to the abstraction and who my target audience is: I've noticed most pixel art tutorials online are basically 'look how I am doing pixel art, try to do the same'. I think this avenue of approach is best left to video tutorials...and that a 'real' tutorial should best approach the fundamentals of the artform from the abstract towards a workable hypothesis...
--- End quote ---
I like this approach. When done right, it can create a deep and engaging dialog. As an example, watch the first 4 minutes of this:
http://www.archive.org/details/halmit1a
This is probably very different from how the average teacher presents an introductory Computer Science course. Any instructor can be comprehensive (this is the history, this is the processor, this is source code, any questions?), but it’s an entirely different thing to present the subject in a genuinely provoking manner. Your opening comparison of pixel art to Go hints at that style of teaching.
--- Quote from: Helm ---...Ostensibly, this has with one leg on academia and one leg on practical application and I will not say this might not be a bit troubling...
--- End quote ---
There is an issue of straddling the gap between the theoretical and the pragmatic. Theory is fine as long as there are some concrete exercises to complement the hand waving. The video lecture above is meant to be used along side the SICP textbook, which provides specific exercises to work through.
Exercises along the lines of “here is an image with some serious issues, address them based off of what you just learned” can solidify the material.
I also like Abelson’s comment that Computer Science is not really about computers in the same way that Biology is not really about microscopes and petri dishes, or in sense that Geometry is not really about using surveying instruments. He then argues that the essence of Computer Science is really about formalizing intuitions about process.
This sentence especially struck a chord with me:
--- Quote ---...when some field is just getting started and you don’t really understand it very well, it’s very easy to confuse the essence of what you’re doing with the tools that you use.
--- End quote ---
In the same way, pixel art is not really about dithering, or banding, or anti-aliasing, or even individual pixels. I wish I could come up with a nice short sentence that sums up what I feel is the true essence of pixel art, like Abelson did with Computer Science. Something that explains a certain aesthetic and process of problem solving. One that is both singular and holistic; both minimal in implementation and maximal in application.
If Pixelation gets its own tutorial, I hope it is written in a way that can instill these values for use in other art forms.
I’ll try to amass more concrete feedback later.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version