AuthorTopic: GR#235 - THE CLUSTER STUDY THREAD!  (Read 33473 times)

Offline Helm

  • 0110
  • *****
  • Posts: 5142
  • Karma: +0/-0
    • View Profile
    • Asides-Bsides

GR#235 - THE CLUSTER STUDY THREAD!

on: July 20, 2013, 09:04:49 am
This is a critique thread, but not like the usual ones. Here we will post art (either ours or not, respectfully) where we marvel at great cluster composition and/or offer better solutions for problems in the clusters. This thread is not about animation, palette choice, style or whatever else, it's strictly about great use of clusters.

For those that do not follow the Ramblethread, a brief definition of "cluster"

  • A group of pixels (more than 1 pixel) of the same color, making a shape.

You can call it whatever you like, pixel group, brush stroke, pixel shape, pixel form, whatever. Just as long as we know what we're talking about.

To understand what makes a good cluster (according to me, at least) either read the Ramblethread or follow this thread as it progresses and you'll figure it out. The purpose of this thread is to raise awareness of solid cluster usage for new artists coming to pixel art and old artists that want to branch out in what is often called 'new school' pixel art technique.

I'll start out with an examination of a sprite from Shin Megami Tensei for the SNES, which is from an age long past, decades before any 'new-school' even existed, or even the term 'pixel art' was floated.



This is such a wonderful sprite. I'm going to focus on a few well done situations and then I'm going to do an edit on a few problematic bits. After reading the 'positives' below, I'm going to ask you to be proactive: study the sprite in your image editor and try to fix as many cluster issues as you see and remove banding. Then compare to my own edit to see what we did differently and why.



A. Although the embroidery bands with the outline of the fabric, I think it's a fair trade because the fabric has great flow and the embroidery really adds to the holy warrior aspect of the sprite. The fabric convincingly is one three clusters. One is the big 'triangle' in the front with a taper towards and under the legs where just one single line width edge conveys the end of the shape. The complicated shadow under that single line width edge is another cluster, which bands a bit and can be cleaned up. The third cluster is the shadow of the fabric (right under the green "A") which also could be done with better pixel tech where it buffers into the primary cluster, but it's still very good.


B. The main thing I love about this sprite is the armor. Given the few colors used, there's some broken outlines here (right over the "B") which I can live with because the mind fills in the gaps and can see it's a continuation of the shadow of the kneepad cluster made out of three pixels above and to the left of the B. There are other solutions to this, but I can roll with that.

Right under the B we have a pretty brave cluster combination. The highlight of the edge of the armor meets the shadow of the whole kneepad, which is a triangular cluster with a strong vertical line. I just adore this sort of solution and do it in my own pixel art all the time. Strong and immediately visible for what it is. The boots of the sprite do not fare as well and are a complicated jumble of shapes, I'll address that in my edit.

C. This is another armor edge which I love, look especially how the 45 degree cluster tapers with a darker red to round out the armor. Again, this sort of thing has been influential on me. I can even give a pass to the single red pixel right above the 'C' because the mind can see it's part of the buffer between the shadow cluster and the highlight cluster on top. There are more elegant solutions, still.

D. This whole thing has some banding, but it's been admirably dealt with on the whole. We have here the continuous, flowing and bending single cluster of the back of the loincloth fabric, and ON TOP we have two stands of clothing flowing over it. As you can see again there's a single pixel line going through them, darkening and lightening itself according to its volume and a lightsource. The only thing to do over that is to remove some banding and make the bends even sharper, I would say.

E. Again, armor and edge highlight. A foreshortened bracer works pretty well visually, I just love it. Only problem is an unfortunate vertical band where the edge highlight meets the shadow of the bracer. Also note that the two edge highlight clusters are the exact shame shape, which can be debated as a good or bad idea.

I am going to list some problematic areas now in vague terms and you should look to address them before reading onwards:
  • open palm and hand over the wing
  • the hair
  • single pixels
  • sword banding
  • boots
  • armor faux dither


Alright, got all that? Here's my take:



There's still problems, but I hope some of my changes make sense.



Now, what to do in this thread:

1. Detailed critique as in the above
2. Just post images with great clusters for us to debate on and study
3. Ask questions to disambiguate the term or various issues surrounding it
4. Post your own images just for cluster edits and critique
5. Debate theory
6. Have fun and learn and rub the bellies of pugs and french bulldogs

Offline hapiel

  • 0010
  • *
  • Posts: 138
  • Karma: +0/-0
  • Wannabe game artist
    • hapiel
    • http://pixeljoint.com/p/9092.htm
    • View Profile
    • OpenPixelProject.com

Re: THE CLUSTER STUDY THREAD!

Reply #1 on: July 20, 2013, 12:01:01 pm
I read too little from the ramble thread and obviously I have no clue what we are talking about here, but I thought I could at least give editing a shot.


I think I made some improvements.. Right?

Also Helm, is that additional sel-out that you did? On the knee and sword? I am confused.
Come check out the OpenPixelProject!

Offline Mr. Fahrenheit

  • 0010
  • *
  • Posts: 327
  • Karma: +0/-0
    • View Profile

Re: THE CLUSTER STUDY THREAD!

Reply #2 on: July 20, 2013, 01:46:15 pm


I mainly tried improving some of the clusters I guess. I also demonstrated my unwarranted hatred for single pixel lines in such places as the fabric and most of the highlights. I didnt really work on the wings or lower legs too much though.

Edit:



A second edit. I removed the free flying fabrics because I didnt really get what they were doing :/ . I messed with the back legs some as well
« Last Edit: July 21, 2013, 12:42:17 am by Mr. Fahrenheit »

Offline pistachio

  • 0010
  • *
  • Posts: 439
  • Karma: +4/-0
  • Mostly lurking
    • http://pixeljoint.com/p/125138.htm
    • View Profile

Re: THE CLUSTER STUDY THREAD!

Reply #3 on: July 20, 2013, 02:38:47 pm
Here's my take:



Aimed for value and color adjustments, simplifying/amplifying light on dark contrast, etc. Tried ways to give it more perspective and make it read at a glance. I'd like to do that to more of the SMT sprites if I have the time, could get interesting results. Been thinking of doing it for years actually, but skill-wise probably couldn't do it justice until recently.

Obligatory cluster adjustments on face and hair, and redone sword. Thought the ribbons positions was a bit monotonous, so repositioned. Most light and dark adjustments took place around there.

Certainly not as drastic as planned and a bit rough where it is, thus effectively WIP. I'll focus on the wings in a second edit if that can be put out and reviewed. Maybe change the angle of the head towards us, making it more engaging. Even bigger hand, maybe ribbons towards the viewer--why not exaggerate perspective? More drastic and complex stuff overall, but just putting this out for now, as that's gonna take a while.

EDIT: Ed. II added, more cleanup, brought back some lost definition in spots like the loincloth and leg armor, did work on the wing, made hand bigger, more hue blending. Tried out a solution for foreleg with more contrast similar to Helm's, basically outside the bounds of the game aesthetic though.
« Last Edit: July 21, 2013, 02:52:12 pm by pistachio »

Offline PixelPiledriver

  • Moderator
  • 0011
  • *
  • Posts: 951
  • Karma: +6/-0
  • Yo!
    • View Profile
    • My Blog

Re: THE CLUSTER STUDY THREAD!

Reply #4 on: July 21, 2013, 02:27:58 am


Kinda messed up in some spots.
I'll come back to this.
« Last Edit: July 21, 2013, 02:29:49 am by PixelPiledriver »
And knowing that it is, we seek what it is... ~ Aristotle, Posterior Analytics, Chapter 1

Offline Mr. Fahrenheit

  • 0010
  • *
  • Posts: 327
  • Karma: +0/-0
    • View Profile

Re: THE CLUSTER STUDY THREAD!

Reply #5 on: July 21, 2013, 03:17:42 am
I love the aura! Very reminiscient of old religious paintings, which I'm sure you knew.  :D

Offline Helm

  • 0110
  • *****
  • Posts: 5142
  • Karma: +0/-0
    • View Profile
    • Asides-Bsides

Re: THE CLUSTER STUDY THREAD!

Reply #6 on: July 22, 2013, 07:30:23 pm
Wonderful edits and comments. Now let's keep moving, find a different piece of art with great/interesting clusters to observe or mess with!

Offline Mr. Fahrenheit

  • 0010
  • *
  • Posts: 327
  • Karma: +0/-0
    • View Profile

Re: THE CLUSTER STUDY THREAD!

Reply #7 on: July 22, 2013, 07:39:12 pm


I think this one is really cool. It's by Blackbeltdude. http://www.pixeljoint.com/pixelart/58859.htm

Offline Cure

  • 0011
  • **
  • Posts: 561
  • Karma: +0/-0
  • Pixel Power
    • View Profile
    • Possum Art

Re: THE CLUSTER STUDY THREAD!

Reply #8 on: July 22, 2013, 11:34:25 pm
Not a very good cluster study imo, there really aren't that many cluster relationships to manage. Despite its size it's mostly giant clusters defining just the major features, anything that looks 'vectory' is probably not ideal for cluster focus. Think about more compact images that try to describe maximum detail in minimal space.

Example: (by thernz)
« Last Edit: July 23, 2013, 12:01:14 am by Cure »

Offline Mr. Fahrenheit

  • 0010
  • *
  • Posts: 327
  • Karma: +0/-0
    • View Profile

Re: THE CLUSTER STUDY THREAD!

Reply #9 on: July 23, 2013, 01:01:41 am
Yeah I agree with you. I didnt think too hard about it at the time  :-[