Ramblethread! A brainstorm approaches!

[Ryumaru]

While that may be true, banding is nothing more than an aesthetic pixel device that has generally been deemed unattractive which is what caused your assault against it in the first place. However, the 50% dither has been proven to be generally attractive. While attractiveness is a subjective thing, large general consensus can hold ground. It obviously took you up until now to discover that the 50% dither bands. Do you now find it less attractive than you used to? While you may not want to admit it I would bet that you do now that you have attached it to the connotation of banding. Your perception of an obviously unattractive element ( common banding) has muddied your perception of something much more harmless ( 50% dither).
Furthermore, I highly doubt that you always saw something slightly disturbing about the dither in the way one would about common banding. Explaining your epiphany as a " thought" shows ( to me, based on what I read) that it was not an aesthetic observation ( 50% dither is unattractive -> why? -> ah, I see, banding!) but ( banding is defined by these terms and has these characteristics -> so does 50% dither -> 50% dither is a form of banding!). In this way you are fixing a non-issue in terms of general aesthetic. If banding was attractive, would you be so quick to eradicate it?

Just as other people can band as they want, so can you avoid it- however id be hard pressed to believe that finding unattractiveness in the 50% dither would come to everyone given enough study of the pixel. While I am at a much lower place in the journey of the pixel than you, I can say I have attained a certain level where these higher pixel orders are not lost on me. I could of course be wrong in my statements as I am not in your head. Maybe you did always find the 50% dither unattractive, but it didn't stop you, or anyone else from using it. The same cannot be said about common banding. If you are trying to fix banding in terms of generally pleasing aesthetics, your quest ends before the 50% dither; if you are trying to fix banding as it's own issue as a characteristic in and of itself ( such as not using a certain color in an oil painting purely for personal reasons) then of course you are to continue.

I apologize if you didn't really want this much discussion on the issue ( or if we are just both at the point where we deem the other " wrong"- a horrible place to be in a discussion), especially since I have not been practicing the medium recently. I have not however, stopped thinking of the pixel- and banding was something I can distinctively remember trying to address in my work before you formed it's current definition and ways to avoid it- and I harken back to those experiences when discussing this. I could tell it was a problem before I knew what it was- and feel that a similar feeling would have existed with 50% dither if it was equally as displeasing as common banding.

[Atnas]

So here's a small selection of patterned dithers that do not contribute to banding. Easing from one to another is bound to prodouce some, but them the breaks.

Hmm, the middle patterns of each row appear to have visible rows, because they have no diagonal quality. The bottom left and top right and left are good. I'd hesitate to use the middle patterns instead of 50% though because it would introduce lines, or... bands... into my transition.

I'm wouldn't be too picky about it in my personal work though, and all are good patterns which I might use, I just don't think visible bands are any better than the super dense (to the point of negligible) banding that is 50% dither.

I think a cluster needs to have banding across a visible length to be considered banding. In the case of 50% dither the fact that the corners are touching doesn't make any one cluster look any more aligned than another, because the clusters are all single pixels and are as offset as possible at their density. Introducing gaps into the mix to try to separate the corners just introduces clusters and potential rows or columns which are worse bands than any 50% could ever be.

Granted, you're talking about the

corners in B, but if that middle line was only a pixel long it would just connect the two lines in the most efficient way possible, with a single pixel and I don't think that counts as banding in any sense I understand it. If the middle line was 2 pixels long then I'd say it was banding, but 50% isnt double pixels, it's as meshed and integrated as you can get. If it was 50% dither with doublewide pixels, yeah, that's banding. But not with squares.

Interesting stuff though and I kinda rambled without a solid point, but it's the ramblethread, right? Not discouraging the notions you're bringing up, just offering my own reaction to them. Whether I disagree with them or not is irrelevant, exploring pixels like this is EXCITING. :'D

[Helm]

It obviously took you up until now to discover that the 50% dither bands. Do you now find it less attractive than you used to?

Chicken and egg is obfuscating in this case. It does not matter. I cannot rely on my memories, perhaps I looked at bandy pixel art when I was 17 and I thought it looked fine. Perhaps something felt off back then too, who knows? I can only go with recent memory and what I feel now. And yes, if you think about something in certain terms, facets of it slowly become clear. Practise and theory together urge me along in a direction, in which, possibly, 50% dither is something I'll be avoiding from now on. It's interesting to me that you feel the need to protect or stand up for the earliest dither. I am not insulting or assaulting it, it isn't such a case as with 'selout' for example, the dither has merit (obviously) and I'm not going to go on a crusade to eradicate it from new-school-pixel-art. It's just observations I make on my own path through the art, that's all I've been doing in this thread.

Atnas, your theory sounds about right that for something to be banding it needs to be more than a single pixel as a cluster. However zoom in on some art and look at dither-aa clash, then even single pixels touching a shape can band and distort the contour. And also, the more you zoom in and the more you can see 50% dither pixels, the more they're segmenting the resolution. It's just an interesting thing.

[Ryumaru]

As I saw it, the beginning of this thread dealt with the minute pixel orders that bring things to the next level of professionalism and aesthetic solidity ( especially with the pixel perfect line examples as well as the balance of resolution disguising and representation) and this seems to gear itself as to what is commonly considered " good practice" in pixel art. So many problems arise when we work with something subjective ( such as the best way to work with a medium) but my defense lies with the thought that the 50% dither is not a poor practice like common banding is, and should not be treated as such. Many oil painters will swear away from the use of black paint, saying that it " deadens" the color and that complimentary mixes should be used to make neutrals instead. What they don't realize is the very style that seemed to advocate such color practices- impressionism never condemned the use of black pigment in this way. The real problem that falsely appears to arise from the use of black ( also prevalent in novice pixel art) is linear ramps that don't accurately change chroma or in the case of lighting and subsurface scattering on skin, hue.

This example is to say that as far as what is generally good pixel practice I believe you are deeming all black pigment ( banding) as poor practice when it is really only the linear ramp ( common banding ) and not the 50% dither ( using black pigment appropriately) that is the problem.

[Helm]

I said there are drawbacks to 50% dither, not that it's all drawbacks.

[ptoing]

^
That I can subscribe to 100%.

Here is an example of where 50% dither (or any kinda dither, depending on how it hugs the edges of shapes or lines) can be problematic. It's something that I coined around 2006 as dither to aa clash, short ditaac.



You can see how in 1B the dither creates antialias between the main red shape and the black and white shapes, thus making the shape look more jagged than wanted.

In 2B the dither fill was made after actually aaing the shapes, as in 2A, and finally in 2C I made some adjustments to the dither at the bottom where the shape is aaed toward white so that there is no dark red of the dither next to the lighter red of the aa to white.

This kinda stuff is mainly a problem when working with limited palettes. Another problem that comes with 50% dither is a style one. Back in the day people pretty much only used dither for 2 reasons. 1) Colour restrictions and 2) On CRT screens dither, esp 50% dither is virtually not perceptible as 2 separate colours and thus the problems of both banding and ditaac become smaller. But nowadays people often use dither because they think it looks cool. It can, but not always does imo.

[Ryumaru]

I said there are drawbacks to 50% dither, not that it's all drawbacks.

Given a small enough area, there are drawbacks to any particular pixel device when battling between good pixel practice and representation of form through squares, no? While banding can only be right if it is accurately describing details ( think occurance of banding in scaled photos) 50% dither's jarring problems only arise on edges such as ptoings example of ditaac.

[Chris2balls]

I hope you don't mind the bump, but I've been trying to find dither patterns that work on edges and this is what I got:

I really like the "corner clusters", I feel that they have a lot of potential.
If you have any more patterns that work, I'd be glad to see them.

[Helm]

That's very interesting. Thanks for posting. It's complex, having to think of what the form you're patterning over is and how its edges go and picking a good pattern related to that on top of everything else.

[Helm]

...Look at these beautiful clusters. Only the 'R' in "Gear" has some banding on the curve that could be eliminated. Also a couple of black pixels where the greendither meets the hair could be removed to the benefit of the image, but whatever. So nice and sharp and well-thought out. Some pixel artist knew their stuff in TOSE.

[9_6]

Isn't the Split in the Es and As and the bright outlines on top of and beneath the letters banding or am I missing something?
That's picture perfectly the example you gave about diminishing "infinite resolution".

There's also 45° anti aliasing on the G which, I guess, works if they tried to go for a glow effect but that is used way too inconsistently to be considered intentional.
Some parts of the logo are blurry, others are crisp to the point of having jaggies like for example the first A.

It seems weird that you who can't not see banding would praise this or am I just grossly misunderstanding the concept of banding here?
That wouldn't surprise me, actually.

[Helm]

Yes it is, but they don't want to convey a sharp edge there, but rather that gradiation. With the palette and the res, that's the best they can do. I only begrudge banding when it's avoidable.

[9_6]

Yes it is, but they don't want to convey a sharp edge there, but rather that gradiation.

I'm looking at hi res versions of that logo and they are all sharp without any gradients on them so why do you think that is intentional?
Is it because of the impact a single pixel has at that resolution that you assume such a mishap can only be intentional given how deliberately everything else has been placed?
Well that doesn't explain the inconsistencies in other places I described earlier.

It makes more sense to me that this is sloppy anti aliasing so that the gradiation is in fact a subpixel.

[rikfuzz]

There's probably brand issues if you start pushing the logo around into your pixel grid to that extent, I can't see any way you could avoid at least some of that antialiasing.

[Helm]

Yes, I took it as a style choice to make the vertical aspect of the logo a bit blurred while keeping the horisontal sharp, and this is the way the artist could do it with the means they had. Look at the cloud shapes, how almost all of them are good clusters, how there's little to no noise and how he buffers from the cloud color to the sky color  selectively with the middle shade, not everywhere making it a blurry mess. Look how the foliage (mostly) avoids banding well and especially look at how sharp and crisp the figure is.

[9_6]

And then look how everything you just described just makes the beginner mistakes and irregular polish on the logo seem all the more jarring.

[Helm]

I don't see it as such a big deal.

[Ryumaru]

I downloaded metal gear solid just for the art. Gameplay was amazing too ( until my emulator stopped saving progress). I don't want to say ahead of it's time, but it was definitely a great hybrid of being informed of higher resolution console graphics and consolidating it into the GBC screen with extremely solid pixel tech. Looks better than some GBA games. I don't have any issue with the logo either, It's easy to tell it was done with intent and it's not glaring anyway.

And those cutscenes too... Loved the inky style- the game had such a great aesthetic.

[Helm]

I've been rereading this. It needs a *lot* of cleanup. I hope people can get through the strange writing style and the hanging metaphors and stuff to the meat of things. What I wanted to get to is that I was discussing with my friend who doesn't know much about pixels about what I've tried to contribute to the form and the two things I sketched out in paper for her to understand where about perfect lines and how to clean up cluster contours, and about banding avoidance. She got both very easily. These two things would be first and most important lessons I would teach a pixel newcomer. I think every pixel artist that wants to progress should take small photos of things and try to abstract them first in 1bit and then in 2bit to 1. learn how to make good lines and 2. learn how to make lines touch without too much banding. This knowledge can be systemized. If I weren't working two jobs, I would create a small curriculum and keep a repository of how people 'solved' the same set of images. Perhaps someone else wants to step up. Something like a high contrast human face to be abstracted in 16x16, a telephone, a car, a baseball, stuff like that, simple and iconic and useful. Not an oil painting.

[Grimsane]

I'd volunteer to assist with that,

something along these lines for the series of images and examples of the task itself to outline?




probably without the animation though, I just couldn't help myself 

[Helm]

Yes, exactly like that, though I'd like to see 1bit rendering, not silluette.

[Grimsane]

more like this?


and I'll start hunting down the relevant images and collate them here for feedback, do you think it should be providing only one rendered example like this? and leave the rest so there is no influence on their approach?

and 16x16 >24x24 >32x32 is probably a great exercise, because to be fair 16x16 it's incredibly difficult and in some cases impossible to get any dicernably readable detail of some subjects in 1bit, but encouraging them to try, then increase their canvas for the attempting to further define the objects. but it definitely has merit trying to fit that detail in, because you push the limits of pixels and cement the boundaries of expression in your mind, and then with the 2 bit you reinforce how those 1 colour cluster boundaries can be overcome with more colours.

and well yeah I just think incrementally increasing resolution would be a really good part of the exercise and certainly not too much to throw at potential participants.

[Carnivac]

Old style telephones in low colors reminds me of a prop in a 4 color hack n slash exploration platform game I was trying to develop about 6 or 7 years ago. 


Ok it's not 2 color but it is small.    I'll be quite now.  I'm not even quite sure what this thread is about. 

[Grimsane]

  that looks interesting, and the phones read quite well, not sure about that background purple and it's effect on readability

heh thought this was a good piece to use to explore gameboy style 2bit, I've read quite a few references about sprite and background layers having the possibility for adjusting brightness levels per layer so did it like that too, although it might've only been 2, I made 3 in my experiment. Anyway, here:

[Carnivac]

I've no problem with the purple but maybe I'm used to it.  Also it's designed for a system where all 27 available colors are roughly equal to all RGB values of 0, 127 and 255 and those 3 of the 4 colors per screen (game area as the status bar could have it's own palette) often changed from screen to screen (flick screen rather than scrolling and the fourth color is always black) to add a bit of variety in atmopshere.  Certainly looks alright on the actual machine I was developing it for anyways and in the correct low resolution.  I kinda miss that project.   There's a full in-game shot of it in my portfolio here on this very site. 

[Helm]

God, the 1bit version is the most difficult. What part of the reference to change for readability, how to convey volume versus detail using only the same tool. Pixelling is hard. 24x24 seems to be a sweet spot for single item renders. It's a good idea to go from 16x16 and up I think, even if the 16x16 piece doesn't work on its own.

By doing this and comparing it to your version I've realized a few things about your technique that need work. Sorry if it's unwarranted critique. Forget about breaking up the highlight line at this resolution. It conveys sparkliness to some extent, I guess. But at the detriment of everything else. Leave the sparkliness for when you have more colors, I would say.

Yes, this is a good path. If you want, go for it, Grimsane. And thank you for the proactivity.

[Grimsane]

No, not at all, thanks for saying so I was wondering the same, I think at 32x32 it works, at 24x24 it's questionable, and at 16x16 it's a really bad approach, I think also having a flexible 1bit colour range is hugely beneficial, and I was approaching it in negative space due to the black nature of the phone, I guess that isn't necessarily the best option at lowest resolution, but admittedly it's not something I've really tried, and I didn't spend too long doing them I'm sure with more forethought some things would become apparent, I actually only added that effect to the 16x16 ones at the last moment to make it match the larger icons (in hindsight rather foolish of me) if we were recommending how to approach it, I think doing them rather swiftly and doing a large amount and then going back and analysing the early attempts with a fresh eye and more experience garnered since would be some good advice, if participants spent a long time trying to perfect their first attempts they might burn up motivation before moving on to do the other images or resolutions, so it'd be a decent piece of advice to place in any forewords/activity brief

and I was pondering whether an alpha background/uncounted background tone might be useful even when subtle it helps separate the object from the background and allows readability of negative space like between the receiver and the cradle. while the brightest tone isn't touching the border and it's conceivable to use it as a background fill I think the background tone is something worth mentioning in the brief, it'll open up the rendering possibilities, and as Icons it seems acceptable

Going to search actively now, and probably post a series of thumbnails of the images. Because if the thumbnails read well enough we'll know the image is good for the task

* Here's a start, I can create an IMGUR gallery with the full resolution pictures once I am finished and the consensus in approval of the images if that makes it easier

[king_bobston]

@Grimsane:
Did you make an own thread for this?
If yes, where?
And if not, will you make an own thread?
I want to try it but I don't want to derail the "ramble". 

@Helm:
I learned more from this ramble than from many tutorials.
Thank you for your work!

[Helm]

I'm glad if I've helped.